Life Sciences:Biostructures and Radiation

The basis of biomedical research in the Institute of Nuclear and Hadron Physics is its competencefor the generation, handling and detection of electromagnetic radiation in a wide photon energyrange. Thus about 50 % of the resources of the institute are devoted to this field of research. Thework is focussed on three main topics(i) Studies on structural dynamics of biomolecules by means of infrared light,(ii) Radiation biology with soft, quasi-monochromatic X-rays,(iii) In-beam positron emission tomography (PET) for quality assurance of charged hadrontherapy.

Structural dynamics of biomolecules: The function of biomolecules is based on their structuralproperties. Experimental approaches for the detection of protein-conformational changes are ofparticularly interest to understand the complex regulation of biomolecular reactions as well as themodulation of physico-chemical properties of biomolecules of potential use in technical applica-tions. Infrared spectroscopy is the major tool used to characterize the conformation of proteins.One focus of our work is the elucidation of structural changes in G protein-coupled receptors usingthe bovine photoreceptor rhodopsin as a model system. Membrane proteins of this class transmita large variety of cellular signals relevant to pharmacological applications. We have identified theinterdependence of transmembrane structure and surface-conformational changes that determinereceptor stability and interaction with the G protein. These data have direct implications for themolecular mechanisms of receptor activation in other class-I receptors.Another system under study is the outer surface layer of bacteria. We are interested in determin-ing their protein-structural features because no crystal structure is available. Our main goal isto understand the basis for the metal-binding ability of this protein class. Unexpectedly, our IRexperiments demonstrate a distinct increase of protein stability in metal-bound S-layer protein.This result is of importance for the technological application of these proteins in biocatalysis orbioremediation, where heavy metals may be sequestered through binding to S-layers.The IR-spectral investigation has provided new insight into structure function relations in thesetwo protein classes. The identified spectral windows may be exploited to study structural dynam-ics by pulsed IR-radiation, that will be available from the ELBE-FEL. Corresponding calculationson undulators have been performed that address the tuning to different spectral regions relevantto biomolecules. In addition, experimental approaches have been advanced which exploit thehigh average power of the FEL IR-radiation as is the case for photothermal beam deflection andnear-field microscopy. These techniques allow to combine spectral and spatial information onbiomolecular samples.

Cell radiobiology: This activity is devoted to the determination of the relative biological effective-ness (RBE) of X-rays with photon energy in the range 10 to 100 keV for different cell types andbiological endpoints. Since such photon radiation is widely used in diagnostic and therapeuticradiology, a precise knowledge of RBE values and moreover of the RBE dependence on photonenergy is highly desirable. For this, intensive quasi-monochromatic X-rays of variable energy willbe produced at the ELBE superconducting electron accelerator by channeling of the electrons incrystals. The beamline of the channeling source was put into operation and first measurementsof channeling radiation produced in diamond crystals have confirmed the prediction of channelingphoton intensity. In preparation of cell irradiations at the ELBE beam, a new laboratory with aconventional highly stable X-ray tube has been established in the ELBE building and will be usedfor reference irradiations. The dosimetric techniques have been refined and applied for precisedosimetrical characterization of the reference X-ray tube. This also involves the development ofadapted algorithms for the Monte Carlo program AMOS, which will be applied for exact photon

45

and electron transport simulations down to cellular dimensions. Furthermore, the cell irradia-tion studies at conventional low-energy X-ray tubes have been continued. The results of theseexperiments revealed a strong RBE dependence on biological endpoint. RBE values obtained fordifferent tube voltages indicate a RBE dependence on photon energy which has to be furtherinvestigated.

In-beam PET for quality assurance of charged hadron therapy: During the three therapy beamtimes of the year 2003 45 patients have been treated by means of intensity modulated carbonion portals at the German Heavy Ion Therapy Facility. Meanwhile carbon ion therapy is offeredand accepted as routine treatment for chordomas and low-grade chondrosarcomas of the skullbase. Locally advanced adenoid cystic carcinomas and sacral or spinal chordomas and low-gradechondrosarcomas were treated in clinical phase I/II studies. All these treatments have beenmonitored by means of in-beam PET. Due to refined algorithms and the installation of state-of-the-art computing hardware the results of an in-beam PET-scan usually become available in lessthan half an hour after finishing the therapeutic irradiation. If inconsistencies to the treatmentplan are observed, the application of our algorithm for a PET-data based calculation of localdeviations from the prescribed dose provides essential information to the radiooncologists for theirdecision on an optimal continuation of the fractionated irradiation. Besides the clinical in-beamPET extensive work on the further development of the method in preparation of its introductionin clinical hadron therapy has been performed: (i) It has been experimentally shown that novelPET detectors consisting of the inorganic scintillator lutetium oxyorthosilicate and avalanchephotodiode arrays can be operated in-beam without degradation of their imaging properties. (ii)A comprehensive experimental study on the feasibility of in-beam PET for quality assurance ofproton therapy has been completed with promising results. (iii) A software tool on the basis ofthe simulation code FLUKA for predicting the positron emitter distribution from the treatmentplan for all therapeutically relevant ions with atomic number 1 ≤ Z ≤ 8 is being developed. (iv)A Monte Carlo study on the potential of in-beam PET for the quality assurance in photon ther-apy indicated that for electron linacs with beam energy above 20 MeV such a technique is feasible.

Structural Dynamics of Biomolecules

◦ Rockefeller University, New York◦ Universitat Orenburg◦ Universitat Nurnberg-Erlangen◦ MPI fur Zellbiologie und Genetik, Dresden◦ ISAS, Berlin, FU Berlin◦ ENEA, Frascati (Italien)◦ CLIO, Paris (Frankreich)

Cell radiobiology

◦ Klinik fur Strahlentherapie und Radioonkologie, TU Dresden◦ Institut fur Kern- und Teilchenphysik, TU Dresden◦ Medizinische Fakultat, Universitat Gottingen◦ Institut fur Bioanorganische und Radiopharmazeuthische Chemie (FZ Rossendorf)

In-beam PET for quality assurance of charged hadron therapy

◦ GSI Darmstadt◦ Radiologische Klinik der Universitat Heidelberg◦ Deutsches Krebsforschungszentrum Heidelberg◦ Institut fur Bioanorganische und Radiopharmazeutische Chemie (FZ Rossendorf)◦ Soltan Institute for Nuclear Studies, Otwock-Swierk, Poland◦ National Institute of Radiological Sciences, Chiba, Japan

46

Structural Stability and Reactivity with Heavy Metal Ions of S-Layer Protein fromBacillus sphaericus

O. Savchuk, K. Pollmann1, J. Raff

1, J. Philipp, S. Selenska-Pobell

1, K. Fahmy

S-Layers are two-dimensional protein arrays that formthe outer surface of a variety of bacteria. They haveattracted attention by their ability to bind heavy metalions and to allow metal nanocluster formation with po-tential applications in bioremidiation and nanoelectron-ics, respectively (see e. g ref. [1]). A strain of Bacillussphaericus (JGA-12) isolated from an uranium-miningwaste site in Johanngeorgenstadt, Germany, has beenshown to accumulate a large variety of heavy metalsfrom drain waters. XAFS-studies have shown that phos-phate groups are likely to coordinate uranium (VI) [2].We have studied by Fourier transform infrared (FTIR)-spectroscopy the structural stability and mode of proteinPd interactions, of the S-Layer of B. sphaericus strainJGA-12. This method is particularly suited to observein real time the proton exchange reations between thetitratable carboxylic acid groups of asp and glu residuesand the bulk solution. The IR absorption allows to cor-relate pH and protein secondary structure.

Fig. 1 pH-

dependent IR-

absorption spectra

of S-Layers in sus-

pension prepared

from B. sphaericus

strain JGA-12.

The changes were

induced by dialysis

against phos-

phate buffers of

the indicated pH

using attenuated

total reflection

spectroscopy.

FTIR-difference spectra where recorded with a Vector22 BRUKER spectrophotometer using dialysis-coupledATR-FTIR-difference spectroscopy[3]. pH changeswhere induced by buffer exchange in the upper compart-ment of the ATR-cell. Figure 1 shows the reduction ofabsorption by the symmetric and antisymmetric COO−

stretches at 1400 and 1572 cm−1, respectively, and thecorresponding increase of the absorption of the C=Ostretch of the protonated carboxylic acids upon loweringthe pH. Half of the initial COO− absorption observed atneutral pH is abolished between pH 2.8 and 3.1, demon-strating a remarkably acidic average pKa of about 3.At pH 0.8, the denaturation of the S-layer protein isevidenced by the downshift of the amide I frequency.

Figure 2 shows that metallization with Pd prevents thepH-induced IR-spectral changes. The absorption of theCOO− groups is almost lacking already at neutral pHand at acidic pH almost no protonation of carboxylicacids occurs. Interestingly, at pH 0.8, the protein struc-ture does not show the changes observed in the absenceof metal ions, indicating an increased stability of themetallized S-layer.

Fig. 2 pH-

dependent IR-

absorption spectra

of Pd-metallized

S-Layers in sus-

pension prepared

from B. sphaericus

strain JGA-12.

The data suggest that in the presence of Pd, car-boxylic acid groups become blocked for protonationand may thus be involved in complexing heavy metalions. Carboxyl-mediated metal protein interactions ap-parently stabilize the S-layer structure. An arrangementthat is in agreement with the FTIR-spectral analysis isshown in Fig. 3.

Fig. 3 Model, which shows

different orientation of

the transition moments

of the two COO−modes

relative to neighboring

metal clusters. The

symmetric COO−stretch

is dipole-forbidden, the

anti-symmetric is surface-

enhanced.

[1] M. Sara, U.B. Sleytr, J. Bacteriol. 182 (2000) 859-868

[2] C. Hennig, P.J. Panak, T. Reich, A. Roßberg, J. Raff, S.Selenska-Pobell, W. Matz, J.J. Bucher, G. Bernhard andH. Nitsche, Radiochim. Acta 89 (2001) 625-631

[3] K. Fahmy, Recent Res. Devel. Biophys. Chem. 2 (2001)1-17

1FZR, Institute of Radiochemistry

47

Involvement of the Extreme C-Terminus of Rhodopsin in Coupling TransmembraneConformational Changes to the Cytosolic Surface

Nicole Lehmann and Karim Fahmy

G protein-coupled receptors (GPCRs) reside in theplasma membrane of virtually all eucaryotic cells. Theycouple extracellular signals such as hormones, neuro-transmitters, and physical stimuli to cellular responses.Understanding the molecular basis of signal percep-tion and the ensuing activation of cytoplasic G-proteins,which trigger cell-specific enzyme cascades, is of funda-mental interest in molecular pharmacology. GPCRs me-diate signals for growth, differentiation, and neuronalactivity and are thus prime targets for the majority ofcurrent pharmacotherapeutics. Light perception by thevisual photoreceptor rhodopsin follows the general prin-ciple of G protein-coupled signal transduction and is anintensively studied model system for the elucidation ofGPCR structure function relations. Rhodopsin is theonly GPCR for which crystal structures have been de-termined which allows a structural interpretation of bio-chemical and spectroscopic data. In rhodopsin, it is thephotoisomerization of the covalently bound chromophore11-cis-retinal which induces a conformational change inthe heptahelical transmembrane domain of rhodopsin.This structural alteration can be followed by Fourier-Transform-Inrared (FTIR) difference spectroscopy. Thefunctionally important transmission of transmembranesterical strain to conformational changes on the receptorsurface, where coupling to the G protein occurs cannotbe deduced solely from the crystal structure of the darkstate of rhodopsin. In addition, part of the cytosolicsurface is not resolved. We have studied by FTIR dif-ference spectroscopy the influence of limited proteolysisat the cytosolic surface of rhodopsin on the transmem-brane structure during relaxation of the photoactivatedmetarhodopsin II state. We found that removal of theextreme C-terminus of rhodopsin by trypsin cleavage un-couples surface-confromational changes from the trans-membrane domain around the Schiff base linkage be-tween the chromophore and the protein. In the absenceof the nine C-terminal amino acids the C=O stretchingfrequencies of protonated carboxylic acids in the vicin-ity of the Schiff base become stabilized in the secondsto minutes time range, whereas surface-conformationalchanges become accelerated. The stable C=O stretchingmode at 1750 cm−1 evidences a stable H-bond particu-larly to Asp83. The FTIR results indicate that a H-bondnetwork along helix 2 connects the transmembrane do-main with the extreme C-terminus at the cytosolic face.Comparison with the spectral changes evoked by chy-motrypsin and papain cleavage suggests that the ex-treme C-terminus interacts with the second cytosolicloop in MII by crossing the cytosolic end of helix 2. Thisinteraction stabilizes the MII structure and can be dis-rupted either by cleavage in the second cytosolic loopor by removal of the C-terminus. The C-terminally me-diated packing of the cytosolic surface is not only re-quired for a stable MII structure but also for binding

of the G protein. In the absence of the C-terminus,binding of a peptide sequence derived from the 11 C-terminal residues of the Gtα subunit destabilizes thetransmembrane conformation of rhodopsin, whereas alocal surface structure absorbing at 1643 cm−1 is stabi-lized by the peptide. This contrasts the general stabi-lization of MII structural features by the peptide whenbound to the intact receptor. In summary, the light ac-tivation generates a transmembrane receptor structurethat does not fully determine the cytosolic conforma-tion. The intrinsic packing properties of the cytoso-lic loops appear to have an important influence on theformation of the G protein-activating receptor surface.

Fig. 1 Upper panel: At-

tenuated total reflection-

FTIR difference spec-

tra taken 15 sec and

2 min after photoacti-

vation of rhodopsin in

urea-stripped disk mem-

branes (which are free

of peripheral membrane

proteins) attached on an

ATR crystal in a bulk

buffer phase (10 mM

NaH2PO4, pH 4.8, 100

mM NaCl, 0oC). Positive

bands belong to the pho-

toproduct state meta-rhodopsin II, negative bands to the dark state of rhodopsin.

Shown are spectra in the absence (solid lines) and presence

of 50% glycerol (broken lines).

Middle panel: Reformation of the 1557 cm−1 absorption of

the dark state over time (decrease of negative band intensity

between 1560 and 1570 cm−1, crosses) is accelerated ver-

sus normal disk membranes, whereas the C=O stretching

modes of internal carboxylic acids between 1700 and 1800

cm−1 is little affected by urea-stripping. It is mainly the

C=O stretch of Asp83 on helix 2 in the MII state which

shows a decrease over time (time course of band intensity

between 1761 and 1750 cm−1, open circles). Glycerol sta-

bilizes these bands (spectra in upper pannel) showing that

surface-conformational changes are coupled to the transme-

brane domain as monitored by the 1768/1750 cm−1 difference

band of Asp83.

Lower panel: Upon removal of the extreme C-terminus (nine

amino acids) by trypsin, the surface-conformational change

causing the 1557 cm−1 absorption is further accelerated,

whereas the decay of the C=O stretching modes of all in-

ternal carboxylic acids is stabilized (band intensity of Asp83

between 1761 and 1750, and of Glu122 between 1745 and

1735, open circles, open triangles, respectively). This shows

that the extreme C-terminus is required for the coupling of

internal conformational changes to those on the cytosolic face

of rhodopsin that is recognized by the G protein.

48

A Scanning Near Field Infrared Microscope Using Etched Chalcogenide GlassFibers

M.Sczepan, J. Martin1

K. Fahmy

Infrared microscopy is a powerful tool for the investiga-tion of biological samples. In contrast to visible lightmicroscopy the absorbance of the samples is high in theinfrared spectral region eliminating the need for stainingor fluorescence techniques. Moreover using infrared lightgives direct access to chemical information i. e. the dis-tribution of proteins, lipids and other biochemical com-ponents in the sample.In conventional microscopy the resolution of the instru-ment is limited by the wavelength of the light and thenumerical aperture of the microscope objective. Sincethe interesting absorbance bands for most biomoleculesare in the 5 - 10µm region and the numerical aperture ofthe reflective microscope objectives is about 0.5 in mostcases a conventional infrared microscope will not be ableto resolve objects in the 5 - 10 µm range such as singlecells.One way to overcome the resolution limits of classical mi-croscopy is to employ near field techniques using smallapertures of sub-wavelength size. In this case the reso-lution only depends on the size of the aperture. Sinceonly a very small part of the light will be transmittedthrough such a sub-wavelength aperture and the effi-ciency of light coupling is strongly dependent of the dis-tance between sample and aperture it is vital to have anintense light source and an effective control system forthe aperture sample distance. Intense light is providedby an infrared laser source while distance is controlledby using an atomic force microscopy technique [1].A near field microscopy setup using the fiber tip illumi-nation scheme has been set up in our laboratory. Lightfrom a laser source (CO2-Laser, FEL or fs-OPA) is cou-pled into a 100µm core diameter chalcogenide glass fiber.A fine tip at the end of the fiber which was generatedusing an etching process [2] serves as a sub-wavelengthaperture. The distance between fiber tip and sample iscontrolled using an atomic force microscopic technique.In a shear force detection scheme the influence of thesurface-tip adhesion forces on the frequency and phaseof the transversal vibration of the tip (generated by apiezo crystal) is used for distance control. The sam-ple is mounted on a xyz-translation stage for scanningand distance variation. A home built refractive infraredmicroscope objective (ZnSe lenses, focal length 20 mm)collects the light from the sample and transports it to an

off-axis paraboloid mirror which focuses the light ontothe detector. A liquid nitrogen cooled Mercury Cad-mium Telluride (MCT) detector is used to measure thesignal.First tests of the microscopic system have been per-formed on various test samples such as narrow metalstrips on glass substrate and liposomes dried on a glasssubstrate. Liposomes are widely used as a model forcell membranes. With infrared near field microscopythe distribution of membrane proteins and other biolog-ically important components in the membrane could bestudied.The fiber coupling and collection optics have been tested.First infrared microscopic tests will be performed witha CO2-laser as a light source.

Fig. 1 Topographic (atomic-force-microscopic) image of a

test sample (dried liposomes on a glass substrate) - instead

of the expected uniform layer the liposomes form islands on

the substrate after drying.

[1] H. Yoshikawa, H. Masuhara, J. Photobiol. Photochem. C: Reviews 1 (2000) 57-78

[2] M.A. Unger, D.A. Kossakovski, R. Kongovi, J.L. Beauchamp, D.V. Palanker, Rev. Sci. Instr. 69(8) (1998) 2988-2993

1TU Chemnitz, Institut fur Physik

49

Exploring the Spatial Resolution of the Photothermal Beam Deflection Techniquein the Infrared Region

H. Foerstendorf1, W. Seidel, J.M. Ortega

2, F. Glotin

2, R. Prazeres

2

Photothermal spectroscopy using an infrared pulsedpump source provides spatial information of a samplesurface. In this model investigation we obtained a spa-tial resolution of approximately 25 µm using a stronglyfocused probe beam in a reflecting scheme.In photothermal beam deflection spectroscopy (PTBD)generating and detection of a thermal wave occur nor-mally in the sub-millimeter length scale. Therefore,PTBD can potentially provide spatial information aboutthe surface of the sample and permits imaging and/ormicrospectrometry. This will possibly lead to a usefultool for investigation of adsorbed species on mineral sur-faces.

In PTBD the thermal wave is generated by intermit-tent laser heating and is detected by a probe laser beam(e.g. a HeNe laser) which is reflected from the surfaceof the sample next to the spot of the incident pumpbeam (FEL). Depending on the modulated intensity ofthe FEL a thermoelastic deformation of the surface isinduced which results in periodically, photoinduced dis-placement of the probe beam [1]. Thus, a different re-flection angle is observed which is monitored by the useof a high resolution position detector placed at a suitabledistance from the illuminated sample (Fig. 1).

As a model compound we investigated the border rangebetween an O+-implanted and an untreated region ofa Ge-substrate by recording time curves of the deflec-tion signal at distinct positions of the surface of the sub-strate using a constant FEL wavelength (11.6 µm). Therecording of the time curves is synchronized to the FELpulses. The deflection signal shows up after a short de-lay time (7 µs) while the thermal wave propagates fromthe incident FEL beam to the probe beam spot. Due

to the different absorption coefficients the deflection sig-nal decreases from the O+-doped region to the nearlytransparent pure germanium. Synchronization of sam-ple micropositioning and data acquisition provides therecording of mapping plots of the sample’s surface (seeFig. 2, inset).In a first approach the spatial resolution obtained wasabout 50 µm [2]. In the present study we improved thespatial resolution by focusing the HeNe laser probe beamin front of the surface of the sample extensively. The di-ameter of the laser spot was about 15 µm at the surfaceof the sample. It was found that the spatial resolution ofthe profile is very sensitive to the focusing of the probebeam. The deflection signal can also be increased by fo-cusing the pump beam which, however, becomes crucialwith respect to the high laser power.The set of the recorded time curves of the deflection sig-nal at distinct sample positions is shown in the insetof Fig. 2. The step width between the distinct sam-ple position was 5 µm. The absorption profile in Fig. 2was extracted from the set of the time curves near themaximum of the amplitude at 10.6 µs. The transitionbetween the O+-doped and the untreated region of thesubstrate can be seen around 1.65 mm relative positionwithin a range of about 25 µm. In the next series of ex-periments we will continue to enhance the spatial resolu-tion to a few microns by utilizing more sophisticated fo-cusing components. Simultaneously, this requires morecomplex samples showing well-defined patterns of dopedregions on the germanium substrates.

AcknowledgementsThe cooperation with the staff of the implanter is grate-fully acknowledged.

DifferentialAmplifier

ReferenceDetector

SpectrometerMode Filter

Focussing

Mirror

He-Ne Laser

Digit.Scope

Sample FEL Macro Pulse Structure

PC

CLIO - FEL Beam

ZnSe - BeamSplitter

PositionDetector

Fig. 1 Schematic diagram of the thermal deflection experi-

ment setup.

Fig. 2 Absorption profile of the border range between

O+-doped and pure germanium (extracted from inset;

t = 10.6 µs). Inset: mapping plot of the set of time curves

of the deflection signal.

[1] M.A. Olmstead et al., Appl. Phys. A 32, (1983) 141-154 [2] W. Seidel, et al., Wiss.-Tech. Ber. FZR-373 (2003) 51

1FZR, Institute of Radiochemistry 2LURE, Universite de Paris-Sud, Orsay, France

50

The Optical Resonator of the IR-FEL at ELBE

W. Seidel, P. Evtushenko, P. Gippner, E. Grosse, R. Jainsch, D. Oepts1, M. Sobiella, D. Wohlfarth, A.

Wolf, U. Wolf, R. Wunsch and B. Wustmann

Intense infrared radiation in the 5 - 30µm range will beproduced in the undulator U27 [1]. In the followingwe outline the actual state of the appropriate resonatorand its main control elements. The optical resonator forthis ELBE FEL has been mounted completely. All me-chanical equipment and optical components (alignmentsystem, interferometer for length stabilization, mirrorwheels and chambers, temperature stabilization of themirrors) have been checked thoroughly and tested forlong-term stability. All components, including the un-dulator and both mirror chambers, have been aligned bymeans of optical methods.To optimize the extraction ratio in the whole wavelengthrange we use 5 mirrors with different hole sizes in the up-stream mirror chamber. The Au-coated Cu-mirrors aremounted on a revolvable holder (wheel), which is fixed toa high-precision rotational stage. A similar constructionwith 3 mirrors is used in the downstream chamber.To ensure the stability of the resonator at wavelengthsdown to 3 µm the mirror angular adjustment requiresa stability in the order of 6µrad. For the initial align-ment of the mirror angles an accuracy in the order of20µrad is required. To achieve this accuracy we builtup an alignment system consisting of two collinear He-Ne lasers. It uses two wall markers and 11 moveableadjustment apertures inside the cavity.A Hewlett-Packard interferometer system is used formonitoring and stabilizing the resonator length (Fig. 1).The interferometer beam is split in two beams (70% and30%). The high intensity beam passes through the same

PC

DC-Mot.

setting �L � cavity detuning

stabilization� (L + �L) < � 0.5 �m

laser

splitter

70%

30%

undulator

outcoupling mirror mirror

interferometer

Fig. 1 Schematic view of the resonator length control system.

Its time constant is one Hz and hence fast in comparison to

the thermal time constant of the resonator.

resonator chamber as the main laser and the electronbeam. The laser passes diagonally from one side of theupstream cavity mirror to a retroreflector on the otherside of the downstream mirror. The low intensity beamis directed to one of the five retroreflectors on the frontside of the mirror wheel close to the working outcouplingmirror. The control electronics for the two interferome-ter arms include a servo system to control and stabilizethe relative distance between the two cavity mirrors us-ing the motorized micrometer drive on the translation

stage of the downstream chamber. There is no activetilt stabilization.We have estimated the maximum intracavity laser powerthat can be expected for the ELBE beam when using thesmallest outcoupling hole. Despite their high reflectivityof more than 99% up to 15W (cw regime) can be ab-sorbed in the mirrors. The entire construction is heatedand the resonator adjustment (length and angle) maybe disturbed by thermal expansion. Moreover that heatload may affect the precision mechanics. At a movableprecision construction in ultra-high vacuum a tempera-ture stabilized system based on water cooling is difficultto be realized. Therefore we introduced special heat iso-lation between the high-precision rotation stage and themirror wheel. The wheel is also made from Cu to re-duce mechanical tension between the mirrors and thesurrounding material. Furthermore, the heat exchangeis improved by a more flexible heat dissipation to theoutside of the vacuum chamber (Peltier element or aircooling) rather than by thermal radiation only.To stabilize the mirror wheel temperature we installeda heater in the center of the wheel. Independently ofwhether the laser is working or not all components areat the same equilibrium saturation temperature slightlyabove the saturation temperature.The temperature behavior of the construction has beenstudied in vacuum by simulating the laser power by a 15W heater at one of the mirrors. After 22 hours the satu-ration temperature of 54 oC was achieved at the mirror.The wheel was 4 oC colder than the heated mirror. Fig. 2illustrates the temperature stabilization. The remainingvariation of the mirror temperature (3.5 oC) changes theresonator length by not more than 1 µm which is muchless than one optical wavelength in general. The inter-ferometer system will correct this change.

0 1 2 3 4 5 6 7 850

52

54

56

58

60

62

tem

pera

ture

[oC

]

time [h]

A A A A A

B B B BCCCC C

Fig. 2 Measured temperature behavior of the mirror (up-

per curve) and the wheel (lower curve) during a simulation

of a working regime of a FEL for the out coupling mirror

wheel (see text; arrows A : laser switched on, arrows B: laser

switched off, arrows C: heating of the mirror wheel switched

on).

[1] http://www.fz-rossendorf.de/ELBE

1FOM Institute for Plasma Physics Rijnhuizen, P.O. Box 1207, 3430 BE Nieuwegein, The Netherlands

51

Effects of Undulator-Field Irregularities

P. Gippner, W. Seidel, D.Wohlfarth, A.Wolf, U. Wolf, R. Wunsch and C.A.J. van der Geer1

In 2000 the magnetic fields of either section of the U27undulator were scanned and analyzed at DESY usinga hall-probe setup [1]. These measurements have beenused to calculate the path of a reference electron (20MeV) through the undulator (fig. 1a). After transport-ing the undulator to Rossendorf and installing the stain-less steel vacuum chamber necessary for the electronbeam the magnetic field has been rechecked by meansof the pulsed wire method [2]. We found significant de-viations from the field measured at DESY (fig. 1b). Wesuppose a displacement of certain magnets since inhomo-geneities in the chamber material could not be detectedby a low-µ permeability indicator. To remove the re-markable displacement of the average electron path wehave shimmed some of the undulator magnets. The re-sults are shown in figs. 1c and d for different gaps.

Fig. 1 Path x(z) of a 20MeV reference electron through the

two units of the U27 undulator (gap g = 15mm) calculated

on the basis of a Hall-probe measurement (a) and second

field integrals J2(z) determined by means of the pulsed-wire

method before (b) and after (c) shimming several undulator

magnets. Fig. (d) shows the field integral measured for the

smaller gap g = 13.5 mm with the same shimmed magnets.

Irregularities in the electron path as seen in fig. 1 affectthe lasing process. If the electrons are displaced from theoptical axis on a part of their path through the undu-lator they experience a lower optical field and the lasergain is reduced. Moreover the resonance wavelength isshifted to longer values since an additional small frac-tion of kinetic energy is moved from the longitudinal tothe transversal motion. Altogether the interaction ofthe electrons with the optical wave depends in a cru-cial way on their path and can not be described by asimple formula in the general case. That is why wehave studied the influence of irregularities in the elec-tron path on laser gain and wavelength by means of the3-dimensional simulation code GPT [1]. Fig. 2 shows anexample where one undulator unit has been studied andthe field strength of 4 magnets has been manipulatedsuch that a part of the electron oscillations is shiftedaway from the axis.

a

�x

Fig. 2 Electron path

with irregularities

The results of such a shift are illustrated in fig. 3. Inde-pendently of the electron energy we can conclude that asignificant reduction of the gain or a shift in wavelengthis only caused if the electron is displaced by more thantwice the amplitude of the regular oscillation in the un-dulator. Irregularities as observed in fig. 1 can easily betolerated by the FEL.

Fig. 3 Gain re-

duction (a) and

wavelength shift

(b) as a function

of the ratio ∆x/a

(defined in fig. 2) cal-

culated for a 20 MeV

(open symbols)

and 40 MeV (full

symbols) electron

pulse. The squares

and circles, respec-

tively, represent a

calculation without

and with beam emit-

tance (15 mm×mrad,

50 ps×keV) taking

into account.

[1] P. Gippner et al.; Proc. of the 23rd Int. Free Electron Laser Conference, II-55, Darmstadt, Germany, August 20-24, 2001;P.Gippner et al., Wiss.-Tech. Ber. FZR-319 (2001) 24; FZR-341 (2002) 27

[2] P. Gippner, A. Schamlott, U. Wolf; http://www.fz-rossendorf.de/ELBE/en/fel/pos.html

[3] M. J. de Loos, C.A. J. van der Geer, S. B. van der Geer,3D Multi-Frequency FEL Simulations with the General Particle Tracer Code, EPAC 2002, Paris, France, p. 849;S. B. van der Geer, M. J. de Loos, The General Particle Tracer Code, Thesis TU Eindhoven 2002, ISBN 90-386-1739-9;Pulsar Physics, General Particle Tracer, http://www.pulsar.nl

1Pulsar Physics, The Netherlands

52

The Role of Resonator Detuning for FEL Gain and Power

R.Wunsch and C. A. J. van der Geer1

In a free-electron laser kinetic energy of a relativis-tic bunch of electrons is transferred into electromag-netic field energy of a copropagating light pulse. Thehigh-frequency accelerator of ELBE produces electronbunches, roughly half a millimeter long (in the labframe), with the repetition rate of νr = 13MHz. To ob-tain a sufficiently strong electromagnetic field, the lightpulses circulate in a resonator. To transfer energy fromthe electron beam to the optical field both pulses haveto overlap in the undulator in a proper way. The lengthLR of the resonator determines the position of the opti-cal pulse with respect to the subsequent electron bunch.The nominal resonator length LR = c/(2νr) guaranteesthat the reflected optical pulse enters the undulator si-multaneously with the subsequent electron bunch andoverlaps with it when moving through the undulator.

It has turned out that the overlap between electronbunch and optical pulse is not optimal for the energytransfer to the optical pulse when both pulses enter theundulator simultaneously. By changing the resonatorlength by a small amount δLR the optical pulse can beshifted with respect to the subsequent electron bunchesand their overlap can be optimized. This procedureis denoted as resonator detuning or desynchronization.There exists an optimal value of the resonator detuningfor which the gain is at maximum. This optimal valuedepends on the electron and undulator parameters. Itis negative and of the order of the wavelength of theproduced light.

Resonator detuning controls the spatial and the spectralshape of the produced light pulse and influences the satu-ration effect. The latter determines the asymptotic valueof the optical power reached after sufficiently many (afew hundreds or thousands) passes of the light throughthe undulator. At optimal detuning the gain is at max-imum but the saturation leads to a small asymptoticpower. Decreasing the detuning the gain decreases butthe asymptotic power increases. Its maximum is reachedclosely to the nominal resonator length where the gainis just about to compensate the optical losses.

Using the code GPT [1, 2] we have simulated the effect ofresonator detuning for one and for two units of the U27undulator of the radiation source ELBE. The electronbunches are 0.7 ps long (rms) in all cases. Fig. 1 showsthe evolution of the optical power for several cases of de-tuning using only one undulator unit. We should men-tion that the calculated build-up time is not realisticallyreproduced by our model. First the spontaneous emis-sion, which is the origin of the amplified radiation, is ar-tificially suppressed in order to overcome the problem ofexaggerated spontaneous emission in FEL simulations.To compensate the suppressed spontaneous emission atthe beginning of the lasing process and to speed up thecomputation time we introduce an additional seed par-ticle. Both measures distort the power increase at the

Fig. 1 Optical power distribution as a function of

the number of passes calculated for various values

of resonator detuning δLR (in units of the light

wavelength λ). Calculation for one undulator unit

(34 periods).

beginning but do not affect the process when the contri-bution from spontaneous emission is negligible.Fig. 1 shows the evaluation of the optical power for var-ious values of resonator detuning. With the nominalresonator length (δLR = 0) the gain is small and thepower develops rather slowly. If the gain is smaller than

Fig. 2 Longitudinal distribution dP/dz of the optical

pulse power as a function of the pass number Npass

calculated for various values of the resonator detuning

δLR, which is given in the boxes in units of the light

wavelength λ. Calculation for one undulator unit.

the optical losses the laser does not start. In the calcu-lation we assumed 5% losses including the outcoupled

1Pulsar Physics, The Netherlands

53

fraction. Shortening the resonator increases the gain.The power grows to a large asymptotic value. Furtherincreasing the resonator detuning increases the gain evenmore but decreases the asymptotic power which is by farnot constant but oscillates with a quite large amplitude.Amplitude and frequency of this oscillations decreasewhen increasing δLR. These power fluctuations are de-noted as limit-cycle oscillations and have been consid-ered in ref. [3] Their frequency is determined by the ratiobetween cavity detuning δLR and electron bunch lengthσz and can roughly be estimated by νosc ≈ δLR

1.5 σzνr. In-

creasing δLR above an optimal value, which is around−0.8 λ in the case of fig. 1, both gain and asymptoticpower decrease. This way the resonator detuning turnsout to be a proper parameter to control important laserparameters, power and spectrum without changing anyof the beam or undulator parameters.The reason for the power oscillations is elucidated inFigs. 2 - 4. Without resonator detuning (upper left pan-els) the power increases from pass to pass mainly in thecenter of the optical pulse and the pulse remains quitenarrow. Accordingly the frequency spectrum is ratherbroad (Fig. 3). If the resonator is shortened the opti-cal pulse moves forward relatively to the electron bunchwhile it grows continuously (upper right panels). Thisprocess continues until the front edge of the optical pulse

Fig. 3 Frequency distribution dP/dν of the optical

pulses shown in Fig. 2.

moves out of the electron bunch and decays. On theother hand the back edge of the optical pulse comes intocontact with the electrons and starts growing so thatthe optical energy swaps from the front end of the op-tical pulse to its rear end. Further increasing the res-onator detuning (lower left panels) accelerates the pro-cess described above (higher power oscillation frequency)while the total power is smoothed. Figs. 2 and 4 showthat growth at one end of the optical pulse and reduc-tion at the other do not completely compensate eachother but may result in oscillations in the total opti-cal power observed in fig. 1. In the frequency spectra(Fig. 3), side peaks appear, while the main peak nar-rows. At δLR = −0.8λ (lower right panels) the sidepeaks have vanished and the spectrum is quite narrow.In this way the resonator detuning can be used to selecteither a particularly short optical pulse, a high opticalpower or a small spectral width.

Fig. 4 The same as in Fig. 2 for two undulator units

(68 periods).

Increasing the number of undulator periods (two undu-lator units, 68 periods) increases the single-pass gain.The laser starts faster. The asymmetry between frontand rear end of the optical pulses is more pronounced(Fig. 4). The main features of resonator detuning, how-ever, are the same a for the shorter undulator.

[1] M.J. de Loos, C.A.J. van der Geer, S.B. van der Geer,3D Multi-Frequency FEL Simulations with the General Particle Tracer Code, EPAC 2002, Paris, France, pp.849;S.B. van der Geer, M.J. de Loos, The General Particle Tracer Code, Thesis TU Eindhoven 2002, ISBN 90-386-1739-9;Pulsar Physics, General Particle Tracer, http://www.pulsar.nl

[2] R. Wunsch, C.A.J. van der Geer, S.B. van der Geer, M.J. de Loos, FZ-Rossendorf, Wiss.-Tech. Ber. FZR-372 (2003) 59

[3] R. Wunsch, C.A.J. van der Geer, S.B. van der Geer, M.J. de Loos, FZ-Rossendorf, Wiss.-Tech. Ber. FZR-372 (2003) 61

54

Design of an Electromagnetic Undulator for a Far Infrared FEL at ELBE

Th.Dekorsy, K. Fahmy, E. Grosse, R. Wunsch

While the U27 undulator will soon produce light in themid-infrared region (5 µm ≤ λ ≤ 25 µm) a new undu-lator for longer wavelengths (25 µm ≤ λ ≤ 150 µm) isenvisaged at the radiation source ELBE.Within this region the FEL constitutes a unique ra-diation source. Radiation quanta of this energy (10 -50meV, 2 - 10THz) are appropriate for spectroscopy oflow energy elementary and collective excitations. Suchexcitations are observed in solid state quantum struc-tures and in complex biomolecules as well. Their studyestablishes the basis for understanding complex phenom-ena in solids and liquids and for elucidating processes inbiological material. Technological and medical innova-tions are the long-term output of such investigation.To produce radiation in the THz region by means of theELBE beam (12 - 40MeV) an undulator with a periodλu of several centimeters is needed. We propose an elec-tromagnetic undulator with λu = 9 cm, 25 undulatorperiods and an undulator parameter up to Krms = 1.8.

Fig. 1 Wavelength λ1 (first harmonic) specified for the two

undulators of ELBE as a function of the kinetic electron

energy Ekine .

This undulator got the preliminary name U90. Fig. 1shows the wavelength region covered by such an FEL.It adjoins to the wavelength of the U27 undulator andallows to produce light up to 300µm. Waves longer than150µm are subject to considerable diffraction losses inthe presently designed beamline to the user laboratories.Above 200µm the transport losses exceed 50%.The diffraction of the light in the resonator increases thebeam radius and reduces the overlap with the electronbeam. Big resonator mirrors, large diffraction losses and

a small laser gain are the result. Approximately above40 - 50µm the losses exceed the gain. These drawbackscan be reduced by means of a waveguide compressingthe optical beam. Two parallel plates are able to guidea particular optical mode without attenuating it notice-ably. To avoid a bulky construction we propose a wave-guide which is restricted to the interior of the undulator.It is 240 cm long, 4 cm wide and 1 cm high. To minimizethe mode conversion losses at the ends of the wave-guide we need bifocal resonator mirrors. The horizontalradius Rh of curvature, is solely determined by the se-lected Rayleigh range of 1 m. The optimal vertical radiusof curvature Rv depends on the wavelength in general.Calculations using the code GLAD [2] have shown thata radius equal to the distance between resonator mirrorand waveguide exit minimizes the mode conversion lossesat λ > 70 µm. Here, the losses per resonator pass do notexceed 1%. At shorter wavelengths the losses grow upto 10% per pass at 25µm. Using a slightly larger radiusof curvature the losses can be reduced to 7 %. At shortpulses (σt ≈ 0.7 ps) the gain is about 15%.Since the number of periods is smaller than in the caseof U27 the spectral width and the average laser ower ofU90 are expected to be larger approximately by a factorof two. The average outcoupled laser power could reachup to 100W.The main parameters of the proposed undulator for far-infrared radiation are summarized in the table below.Details can be found in the internet [1].

Undulator Field period 9 cmNumber of periods 25Undulator parameter 0.4 - 1.8

Resonator Length 1153 cmMirror radius of curvature Rh : 594 cm

Rv : 457 cmWaveguide Length 240 cm

Height 10mmWidth 40mm

Table 1 Main parameters of the proposed U90 undulator

including optical resonator and partial waveguide.

The beam will be outcoupled by a 6mm diameter holein the center of the upstream resonator mirror and thentransported to the diagnostic table in the neighboringcave, as from where it will be delivered to the user lab-oratories using a transport system in common with themid-infrared beam from the U27 undulator.

[1] https://www.fz-rossendorf.de/pls/rois/Cms?pNid=471

[2] GLAD, Applied Optics Research, Woodland, WA 98674, USA

55

Measurement of Channeling Radiation at ELBE

W.Wagner, A.Panteleeva, J. Pawelke, W.Enghardt

The beamline for Radiation Physics at ELBE was putinto operation in October 2003.Research work on the production of channeling radia-tion (CR) has been started in view of its feasibility forapplication as a non-conventional quasi-monochromaticX-ray source to be used for biomedical investigations [1].

The experimental setup consisted in a 42.5 µm thickdiamond IIa crystal fixed on a 3-axes goniometer. Thisdevice was constructed by FMB Berlin and especiallydesigned to meet the rather strong demands for han-dling it under the extremely clean UHV conditions inline with the superconducting electron accelerator.Spectrometry of the CR observed has been carried outusing Si-pin-diodes as well as CdZnTe detectors. Forshielding against the intense background radiation thedetectors were housed inside a 10 cm thick Pb colli-mator with a diaphragm of 5 mm diameter which waspositioned under zero degrees with respect to the beamdirection at a distance of 3.2 m from the target crystal.To avoid a high energy threshold due to the attenua-tion of radiation in air an evacuated (2 mbar) auxiliarypipe was mounted between the Be-window (127.5 µm)of the UHV electron beamline and the detector station.

Fig. 1 Energy spectrum (lower one) of (110) planar CR at

an electron energy of 17 MeV. For clearness, the spectra

measured at aligned and random (grey) crystal orientation

are shifted on the ordinate scale by +1000.

Measurements of planar and axial CR spectra have beenperformed for the zone of the crystal axis [110] at elec-tron energies (Ee) of 14.5 and 17 MeV. For illustration,

the spectrum measured for the (110) plane at 17 MeVis given in Fig. 1 together with the related spectrumof bremsstrahlung which has been registered at slightlynon-aligned crystal orientation and was normalized tothe (110) spectrum at its high-energy tail.The spectral distribution of CR (also shown in Fig. 1)is then obtained as the difference of both these spectra.The measured energy values (Ex) of some prominentCR lines are listed in Tab. 1.

Tab. 1 Energies of prominent CR lines.

Ee/MeV Index Transition Ex/keV14.5 (110) 1–0 15.8

(111) 2–1 11.017.0 (100) 1–0 13.3

(110) 1–0 21.32–1 10.8

(111) 2–1 13.73–2 10.81–0 8.5

(113) 11.217.0 [110] 2p–1s 68.6

3d–2p 44.03p–2s 28.2

During the first run absolute monitoring of the electronbeam current which typically amounted to about 1 nAwas not yet available. The installation of a secondaryelectron monitor (SEM) [2] directly behind the sourcecrystal to avoid electron losses due to the scattering inthe target is still in progress.Some very rough estimation of the CR yield, however,could be carried out by calculating the number of elec-trons from the output of bremsstrahlung registered si-multaneously with CR. Partial integration of the mea-sured spectrum has been performed at an energy largerthan 70 keV where possible coherent contributions fromfree-to-bound transitions are negligible.The resulting yield for the 1–0 transition of planar CRat 17 MeV (Fig. 1) amounts to 0.088 photons/sr e−. Al-though the uncertainty of this estimation is rather high(about 70%) and systematic errors for the time being ig-nored may lead to a substantial underestimation of theCR yield, this value converts into a CR rate of 1.6×1010

photons/s 10% BW if one assumes an average beam cur-rent of 100 µA. By the order of magnitude, such CR ratemeets the requirements for investigation of RBE valuesof soft X-rays on living cells [3].

[1] W. Neubert, W. Enghardt, U. Lehnert, E. Muller, B. Naumann, A. Panteleeva, J. Pawelke, Proc. Conf. Monte Carlo2000, Lisbon 2000, Eds. A. Kling, F. Barao, M. Nakagawa, L. Tavora, P. Vaz, Springer-Verlag Berlin-Heidelberg-NewYork 2001, p. 123

[2] V.V. Morokhovskyi, PhD-Thesis D17, TU-Darmstadt, 1998

[3] A. Panteleeva et al., FZ Rossendorf, Wiss.-Tech. Ber. FZR-271 (1999) 95

56

Establishment of a Laboratory for X-Ray Experiments at ELBE

J. Pawelke, S. Eckert1, W.Enghardt

At the ELBE facility intensive quasi-monochromatic X-rays are produced by channeling of the ELBE electronbeam in diamond crystals [1] in the radiation physicscave (room 112, cf. Fig. 1). The channeling X-ray source(Eγ ≈ 10−100 keV) is planned to be used for radiobio-logical experiments by the Radiation Physics Division [2]and for imaging of phase transition and flow phenomenain liquid metals by the Magnetohydrodynamics Depart-ment of the Institute of Safety Research [3].For the preparation of the experiments including thenecessary detector tests without requiring ELBE beamtime and not occupying the irradiation site in the ra-diation physics cave, a separate X-ray laboratory withtwo conventional X-ray tubes (see Table 1) has been es-tablished in the ELBE building (room 112a, cf. Fig. 1).The high stability constant potential X-ray tube ISO-VOLT 320 HS (Rontgen Seifert, Ahrensburg) is very wellsuited for dosimetrical experiments and will be used asa reference X-ray source for the radiobiological studies,whereas the X-ray unit XS225D OEM (Phoenix X-raySystems, Wunstorf) meets the requirements of featurerecognition down to µm spatial resolution.Important features of the laboratory are: circulating aircooling, argon gas and cooling water (20 l/min, 30◦C)supply as well as access control system. The front wallstands back from the adjacent rooms giving place for theX-ray tube control units (Fig. 2). For radiation shield-ing, the laboratory steel structure has intermediate leadpanels (thickness in the front wall of 28 mm and in theside wall toward the positron laboratory of 100 mm).Perspectively, the use of the quasi-monochromatic X-ray

beam from the radiation physics cave in the X-ray lab-oratory is possible.

Tab. 1 Main parameters of the two X-ray tubes.

Isovolt 320 HS XS225D OEMTube voltage [kV] 5 to 320 10 to 225Tube current [mA] 0.1 to 45 0.005 to 3Maximum anodedissipation [VA] 4200 1400Beam angle [◦] 40 25Anode material W W, CuFocal spot size 4.0 / 1.5 mm ≥ 3 µm

Inherent filtering 7 mm Be 0.5 mm BeMinimum focus-object-distance 13 cm 4.5 mm

Fig. 2 Control unit of the Isovolt X-ray tube in front of the

laboratory.

Fig. 1 General layout of the radiation source ELBE with the radiation physics cave (112) and the X-ray laboratory (112a).

[1] W. Wagner et al., Measurement of channeling radiation at ELBE, This Report, p. 56

[2] W. Neubert et al., Proc. Conf. Monte Carlo 2000, Lisbon, Eds.: A. Kling et al., Springer-Verlag (2001) 123

[3] S. Eckert et al., FZ Rossendorf, Wiss.-Tech. Ber. FZR-341 (2002) 96

1FZR, Institute of Safety Research

57

Dosimetric and Spectral Characterization of an ISOVOLT 320/13 X-ray Tube

J. Pawelke, T.Mikuletz1, A. Panteleeva

One area of research at the new ELBE facility is devotedto radiobiological studies by means of cell irradiationwith low energy photons. For this, the soft X-rays areproduced by channeling of the ELBE electron beam indiamond crystals [1] in the radiation physics cave. Ref-erence irradiations with a conventional tungsten anodeX-ray tube ISOVOLT 320/13 HS will be performed in anearby laboratory at ELBE [2].Absorbed dose will be measured by standard dosimetrictechnique with air filled ionisation chambers (IC), op-timised with respect to dose and energy dependence ofresponse. In order to cover the wide range of photon ra-diation at ELBE (from several keV channeling X-rays upto 40 MeV bremsstrahlung) three different IC (Table 1),connected to an UNIDOS electrometer (PTW Freiburg),are used for high precision dose measurements. Alto-gether two identical systems (3 IC + UNIDOS) are usedfor measurements in both the radiation physics cave andthe X-ray tube laboratory.The dose response of all IC has been measured for thesame conditions of X-ray tube irradiation changing thetube voltage UH from 5 to 320 kV. In agreement withthe recommended energy range of the IC given in Ta-ble 1, the correct dose is obtained by the Farmer IC forthe whole tube voltage range, whereas the soft X-rayIC and the rigid stem IC show correct values only forUH � 100 kV and UH � 200 kV, respectively (Fig. 1).For each chamber type the dose response of both systemswas compared, revealing a 10% dose underestimationfor one Farmer chamber due to lack of recent calibra-tion. Therefore, radioactive control devices will be usedin future in order to control the necessity of IC recal-ibration. The error contribution to the measured dosewas determined to be less than 0.5% from the charge (orcurrent) measurement with the electrometer, 2 % to 5 %from the calibration data depending on the IC, 0.1% to1.5% due to air density correction depending on exper-imental conditions and 0.25% (at 1.75 m distance) to1.6% (at 0.25 m distance) due to positioning the IC ata given distance from the X-ray tube.The homogeneity of beam intensity was measured fordifferent beam filtering and tube voltages. A maximumdose difference of 4.5% was measured within the irradi-

ation field if anisotropic beam scattering in the environ-ment (wall, experimental setup) is minimised by the useof appropriate aperture.Spectral dose distribution was calculated on the base ofspectral photon flux measurement [3] with a high reso-lution 3×3×1 mm3 cadmium telluride diode XR-100T-CdTe (Amptek, Bedford, USA). In order to achieve thenecessary flux reduction, the detector was positioned ata distance of 8 m from the X-ray tube in the radiationphysics cave and the tube was operated at the minimumtube current in order to achieve the necessary flux re-duction. Consideration of detector geometry as well asdetector dead time and efficiency correction results ina reasonable agreement between dose rate directly mea-sured with rather high precision by IC with the less ac-curate dose rate calculated on the base of photon fluxmeasurement (Table 2). However, the dose calculated inthis way is useful in determining a precise spectral dosedistribution if the flux is limited to � 15000 ph/s.

Fig. 1 Dose deviation of the soft X-ray and the rigid stem

IC with respect to the response of the Farmer chamber.

Tab. 2 Photon flux Iγ measured with the detector at 0.1 mA

tube current and corresponding dose rate D scaled to 10 mA

for comparison with IC measurement.

Tube parameters Iγ D [mGy/min]UH [kV ] Filter [ph/s] Detector IC

90 kV no 19500 0.9±0.7 1.1±0.1100 kV 50 µm Mo 15500 0.6±0.5 0.8±0.1160 kV 0.5 mm Cu 27500 1.1±1.1 1.8±0.1200 kV 0.5 mm Cu 41300 2.0±1.8 3.2±0.1160 kV 2.0 mm Cu 7500 0.4±0.5 0.7±0.1200 kV 2.0 mm Cu 13300 0.7±1.0 1.5±0.1

Tab. 1 Main parameters of the ionisation chamber types (PTW Freiburg) used for photon dosimetry.

Chamber type Soft X-ray, M23342 Farmer, M30001 Rigid stem, M23332Sensitive volume 0.02 cm3 0.6 cm3 0.3 cm3

Response 1·10−9 C/Gy 2·10−8 C/Gy 1·10−8 C/GyEnergy range 7.5 keV – 70 keV 30 keV – 50 MeV 140 keV – 50 MeVDose range 3 mGy – 30 Gy 100 µGy – 1 Gy 200 µGy – 2 Gy

Dose rate range 20 mGy/min – 84 kGy/min 0.6 mGy/min – 2.8 kGy/min 1.3 mGy/min – 5.6 kGy/min

[1] W. Wagner et al., Measurement of channeling radiation at ELBE, This Report, p. 56[2] J. Pawelke et al., Establishment of a laboratory for X-ray experiments at ELBE, This Report, p. 57[3] J. Pawelke et al., FZ Rossendorf, Wiss.-Tech. Ber. FZR-341 (2002) 92

1Hochschule Mittweida (University of Applied Sciences)

58

RBE of Soft X-Rays for Cell Survival in a Human and a Rodent Cell Line

A.Panteleeva, W.Dorr1, E. Lessmann, J. Pawelke

Recently, the energy dependence of the relative biolog-ical effectiveness (RBE) of X-rays has become a topicof intensive discussion. In order to draw definite con-clusions, data for different biological objects and end-points in a wide energy range are necessary. An RBEof 25 kV X-rays in the range 1.1 - 1.3 has been previ-ously found for one human and two rodent cell lines byclonogenic survival [1]. Since the X-rays in the energyrange below ∼ 30 keV are applied in diagnostic radi-ology, and especially in mammography, the study wasextended to the human mammary epithelial cell lineMCF-12A. As a further step towards investigation ofthe RBE dependence on photon energy, the clonogenicsurvival after irradiation with 10 kV X-rays was stud-ied for MCF-12A cells as well as for the well-establishedmouse fibroblast cell line NIH/3T3. The culture condi-tions have already been presented for the NIH/3T3 cells[2] and for MCF-12A [3]. The study was performed atthe Medical Faculty of TU Dresden, where the referencesource was a tungsten anode X-ray tube, operated at200 kV with 0.5 mm Cu filtration, resulting in a doserate of 1.2 Gy/min. The irradiations with soft X-rayswere performed using a tungsten anode X-ray tube. Forthe 10 kV X-rays irradiation, no filtration was used andthe dose rate was 0.53 Gy/min, whereas for the 25 kVirradiation, a 0.3 mm Al filter was used, resulting in adose rate of 1.9 Gy/min. For the clonogenic assay, ex-ponentially growing cells were irradiated in polystyrene

culture flasks and afterwards seeded at low densities. Af-ter an incubation period of about 2 weeks, the formedcolonies were stained and scored. In order to avoid addi-tional attenuation in the cell medium, the flasks for theirradiation with 10 kV and 25 kV X-rays were placedupside down in the holder. Since the longest irradia-tion time did not exceed 15 min, no effects of mediumdepletion should be expected. The dose dependencesof the clonogenic survival for the used radiation qual-ities together with the fitted survival curves accordingto the linear-quadratic model S = exp(−αD − βD2) arepresented in Fig. 1. From the resulting coefficients forboth cell lines, presented in Table 1, RBE0.1 at the 10%survival level of 10 kV and 25 kV X-rays of 1.1 - 1.3is obtained, where for both cell lines 10 kV X-rays areslightly more effective in cell killing. The resulting RBEis in good agreement with the data for other cell linesand with the expected photon energy dependence. Wefurther investigated the RBE dependence on the sur-vival level. Considering a 95% confidence interval (CI),the RBE of 10 kV X-rays was found to be significantlyhigher than 1 for surviving fractions in the range 0.03 -0.6, increasing with increasing survival, or, equivalently,decreasing dose. The effectiveness of 25 kV X-rays wasfound to be significantly higher than 1 in the survivingfraction range 0.03 - 0.2, but decreasing with increasingsurvival for both cell lines.

0 2 4 6 8 10

0.01

0.1

1

Sur

vivi

ng f

ract

ion

Dose [Gy]

200kV 25 kV 10 kV

Fig. 1 Mean values for survival and standard errors of the mean of NIH/3T3 (left) and MCF-12A (right) after irradia-

tion with 10 kV (open triangles, dotted lines), 25 kV (open circles, dashed lines) and 200 kV X-rays (filled circles, solid lines).

Tab. 1 Coefficients (± standard error) of the linear-quadratic model and the RBE0.1 at the 10% survival level (± 95% CI)

for survival after irradiation with 10 kV, 25 kV and 200 kV X-rays. Also presented are the coefficients of determination R2.

Cell line Radiation quality α [Gy−1] β [Gy−2] R 2 RBE0.1

NIH/3T3 10 kV X-rays 0.288 ± 0.082 0.048 ± 0.017 0.994 1.34 ± 0.1225 kV X-rays 0.050 ± 0.048 0.088 ± 0.017 0.978 1.25 ± 0.14200 kV X-rays 0.169 ± 0.022 0.035 ± 0.006 0.996

MCF-12A 10 kV X-rays 0.460 ± 0.061 0.034 ± 0.014 0.986 1.21 ± 0.0625 kV X-rays 0.226 ± 0.055 0.078 ± 0.011 0.992 1.13 ± 0.06200 kV X-rays 0.331 ± 0.037 0.034 ± 0.006 0.993

[1] A. Panteleeva et al., Radiat. Environ. Biophys. 42 (2003) 95

[2] A. Panteleeva et al., FZ Rossendorf, Wiss.-Tech. Ber. FZR-319 (2001) 103

[3] A. Panteleeva et al., FZ Rossendorf, Wiss.-Tech. Ber. FZR-372 (2003) 64

1Dept. of Radiotherapy, Medical Faculty, Technical University of Dresden

59

RBE of Soft X-Rays for Micronuclei Induction in a Human and a Rodent Cell Line

A.Panteleeva, W. Dorr1, E. Lessmann, J. Pawelke

The cytokinesis-blocking micronucleus assay is a well-known alternative of the more laborious chromosomalaberrations test. The relative biological effectiveness(RBE) of 25 kV X-rays has been previously determinedfor micronuclei induction in two human and one ro-dent cell lines, resulting in an RBE value of ∼ 1.3 [1].These studies have been extended by the human mam-mary epithelial cell line MCF-12A, since it presents amodel system for risk estimation from mammography,which is performed with X-rays below ∼ 30 keV. Onthe other hand, in order to investigate the dependenceof the RBE on the photon energy, the RBE of 10 kVX-rays was determined for one of the previously stud-ied cell lines, NIH/3T3 mouse fibroblasts. The detailsof the cell culture have been previously described forNIH/3T3 [2] and for MCF-12A [3]. The X-ray tube ir-radiations were performed at the Medical Faculty of TUDresden. A tungsten anode X-ray tube, operated at200 kV with a 0.5 mm Cu filter, was used as a referencesource. The tungsten anode soft X-ray tube was oper-ated at 25 kV with 0.3 mm Al filtration or at 10 kVwith no filtration. The dose rates were 1.2 Gy/min,1.5 Gy/min and ∼ 0.58 Gy/min for the 200 kV X-rays,25 kV X-rays and 10 kV X-rays, respectively. For the mi-cronucleus test, the cells were irradiated in Petri dishes(3 replications per dose point), incubated with cytocha-lasin B for 24 h (NIH/3T3) or 48 h (MCF-12A), fixedand stained with Giemsa. The fraction of binucleatedcells (BNC), fraction of binucleated cells with micronu-clei (BNC+MN) and the number of micronuclei per bi-nucleated cell (MN/BNC) were determined. The result-ing dose dependence for the effect S was fitted to thelinear-quadratic model S = c + αD + βD2.For the RBE of 25 kV X-rays determination, the irradi-ation of MCF-12A was performed on the same passageparallelly for both radiation qualities. The dose depen-dence of BNC+MN is presented in Fig. 1 together withthe linear-quadratic model fit, where the response at zerodose was fixed to the control level. A value of 1.4 wasfound for the RBE, which is in a good agreement withthe previously published values for other cell lines [1].The irradiation of NIH/3T3 with 10 kV X-rays was per-formed in Petri dishes, placed upside down in a holder.Thus the X-ray attenuation was minimised, since onlythe dish wall had to be penetrated by the beam andthe cells were not covered with the liquid medium dur-ing irradiation. However, variation in the dish thicknesscould still lead to variation in the absorbed dose. There-fore, after cell staining, the thickness of the dishes wasmeasured and the absorbed dose for each dish was cal-culated from the measured thickness dependence of dose

rate. The resulting data were pooled and fitted to thelinear-quadratic model, where the β coefficient was setto 0 and the response at zero dose was fixed to the con-trol level. Since the application of high doses resultedin saturation of the response, the values for 4 and 5 Gyof 200 kV X-rays and 2.69 Gy of 10 kV X-rays were notincluded in the fit. The scatter of the dish thicknesseswas found to be sufficiently small to present the data bythe corresponding means and standard error of the mean(SEM) from the 5 experiments, together with the meandose values and the corresponding dose SEM (see Fig. 2).The RBE of 10 kV X-rays for micronuclei induction inNIH/3T3 was calculated to be in the range 1.1 - 1.2.This values are smaller than expected from the assumedRBE photon energy dependence. Therefore, determina-tion of RBE of 10 kV X-rays for micronuclei inductionin other cell lines is planned.

Fig. 1 Dose response of micronuclei induction in MCF-12A

cells, irradiated with 25 kV X-rays (open circles, dashed line)

and 200 kV X-rays (closed circles, solid line). The mean val-

ues and the SEM from 6 experiments are shown.

Fig. 2 Dose response of micronuclei induction in NIH/3T3

cells, irradiated with 10 kV X-rays (open triangles, dashed

line) and 200 kV X-rays (closed circles, solid line).

[1] D. Slonina et. al., Radiat. Environ. Biophys. 42 (2003) 55

[2] A. Panteleeva et. al., FZ Rossendorf, Wiss.-Tech. Ber. FZR-319 (2001) 103

[3] A. Panteleeva et. al., FZ Rossendorf, Wiss.-Tech. Ber. FZR-372 (2003) 64

1Dept. of Radiotherapy, Medical Faculty, Technical University of Dresden

60

Determination of RBE of 25 kV X-Rays for Chromosomal Aberrations inMCF-12A Human Mammary Epithelial Cells

A.Panteleeva, E. Wolfring1, E. Lessmann, J. Pawelke, W.Dorr

2

The relative biological effectiveness (RBE) of soft X-rays has been recently determined by clonogenic sur-vival [1] and micronucleus test [2] for the human mam-mary epithelial cell line MCF-12A. For 25 kV X-rays, theRBE was found to be endpoint-dependent, resulting ina larger value when studied at the level of chromosomaldamage than for cell kill. The micronuclei are formedby a certain kind of chromosomal aberrations (CA) -the chromosome fragments without spindle attachmentorganelles (acentric fragments). During the cell division,some of these fragments are excluded from the daugh-ter nuclei and form the micronuclei, either on their own,or in conjunction with other fragments within the cy-toplasm of the daughter cells. In order to resolve thechromosomal changes caused by soft X-ray irradiation,the RBE of 25 kV X-rays was determined by CA analy-sis.The culture conditions of the MCF-12A cell line havebeen described elsewhere [3]. The irradiation of synchro-nised cells (confluent cultures were maintained for sev-eral days), was performed in polystyrene culture flasksat the Medical Faculty of TU Dresden. As a referencesource, a tungsten anode X-ray tube, operated at 200 kVwith a 0.5 mm Cu filter, at dose rate of 1.2 Gy/min, wasused. The irradiation with 25 kV X-rays was performedwith a tungsten anode X-ray tube with 0.3 mm Al fil-tering at dose rate of 1.9 Gy/min, and the flasks wereplaced upside down in the holder. Immediately afterirradiation, the cells were transported to the cell labo-ratory in FZ Rossendorf for further processing. In orderto inhibit repair processes which could bias the final re-sults, the culture flasks were kept on ice. The cells weresplitted at low density and incubated for 39 h, wherefor the last 3 h of incubation, colcemid at concentra-tion 0.3 µg/ml was added to the medium in order to ar-rest the cells in the metaphase. Parallelly, control of thenumber of cell cycles passed after the irradiation was per-formed by incubation of cells (controls or irradiated with5 Gy) with BrdU. A metaphase suspension was obtainedby scraping the cells from the flask surface. After treat-ment in a hypotonic solution of 10 mM trisodium cytratedihydrate and 32 mM KCl, the cells were fixed in a 3:1methanol/acetic acid solution and dropped onto grease-free, cold, wet slides. The metaphase slides were stainedin Giemsa and observed with magnification of 1000. Theacentric fragments, dicentric chromosomes and centricrings in up to 500 cells per dose point were scored, wherethe dicentrics and centric rings were united in one CA

group of exchange aberrations, one acentric fragmentbeing associated with each exchange aberration. Thedose response curves for the excess acentric fragmentsand for dicentrics and centric rings were fitted to thelinear-quadratic model N = c + αD + βD2, where theresponse at zero dose was fixed to the control level. Fromthe results, presented in Fig. 1, RBEM = α25 kV /α200 kV

was calculated, resulting in a value of 4.4 ± 2.6 for thenumber of excess fragments, whereas for the exchangeaberrations an RBEM of 0.55 ± 0.56. These results, al-though indicating a high effectiveness of 25 kV X-rays,do not allow definite conclusions about the RBE and willbe repeated with better statistics.

Fig. 1 Dose dependence of excessive fragments (upper plot)

and exchange aberrations (lower plot) induction in MCF-

12A cells, irradiated with 25 kV X-rays (open circles, dashed

line) and 200 kV X-rays (closed circles, solid line).

[1] A. Panteleeva et. al., This Report, p. 59

[2] A. Panteleeva et. al., This Report, p. 60

[3] A. Panteleeva et. al., FZ Rossendorf, Wiss.-Tech. Ber. FZR-372 (2003) 64

1Freiberg University of Mining and Technology2Dept. of Radiotherapy, Medical Faculty, Technical University of Dresden

61

Monte Carlo Methods for Electron Transport by Single Scattering Modells

U.Reichelt, J. Henniger1, W.Enghardt

For the investigation of biological effects by irradiatingcells with photons and electrons the spatial and spectraldistribution of primary and secondary electrons, whichare the main reason for damage production, have to bewell known. Therefore, exact electron transport simula-tions down to subcellular dimensions are necessary.Only single scattering models can be used in order toavoid uncertainties as they occur using pure multiplescattering models in small dimensions and thin layers.Adapted algorithms for ionisation and elastic scatter-ing were developed for the Monte Carlo (MC) programAMOS [2], which are efficient enough to handle the largenumber of interactions (≈ 106 for a 100keV electron).If an atomic shell is ionised by electron impact, thesecondary electron spectrum is symmetric as shown inFig. 1. This results from the indistinguishability of bothelectrons and the dependence of their energy values.Only one half of the spectrum is required to describethe ionisation completely.

Fig. 1 Recoil spectrum by ionisation of the aluminium K

shell for an incidence energy of 16 keV

To interpolate spectra for various primary electron en-ergy values E0, the total recoil spectra have to be sep-arated into the ones of the single electrons. Thereforea power extrapolation is used. Its parameters are de-termined by using the data points with highest energyvalues E. After the interpolation to the requested en-ergy value E0 the resulting single electron distributionis summed with the same but mirrored one to the sec-ondary electron spectra. Of course the influence ofthis summation is only important in the middle of thespectrum. But also these quite rare events significantlycontribute to the slowing down of an electron concerningthe relatively high energy loss.Further on an algorithm to handle the elastic scatteringof electrons was developed. The scattering angle dis-tributions of electrons are extremely anisotropic. The

probability density f of the example shown in Fig. 2varies over eight orders of magnitude. With increas-ing energy the forward scattering dominates more andmore. That leads to the lateral broadening of e. g.an electron needle beam. But also the very rare hardscattering events are of importance for problems likebackscattering at material interfaces and in very smalldimensions, as there are at cell irradiation experiments.Fig. 2 shows the exact reproduction of the angle distri-bution by AMOS.

Fig. 2 Distribution of elastic scattering angle θ for 64 keV

electrons in aluminium

To verify the developed MC algorithm it is comparedwith experimental and simulated data from the liter-ature. Fig. 3 shows transmission spectra for two alu-minium layers of different thickness. The results of bothsimulations, AMOS and PENELOPE [3], fit very well.But in AMOS a much smaller binning could be usedbecause of its high processing speed. The discrepanciesto the data by Thomas [3] are conjectured to have theirreason in less well known experimental conditions.

Fig. 3 Simulated transmission spectra of 20 keV electrons at

1 µm and 0.3 µm in comparison with results by experiment

and the MC program PENELOPE

[1] Henniger J., AMOS - ein multivalent nutzbares Programmsystem zur Beschreibung von Strahlungstransportproblemen,Strahlenschutz: Physik und Messtechnik, Band 1, Verlag TUV Rheinland (1994) 145-150

[2] Perkins S. T. et. al., Evaluated Electron Data Library, Lawrence Livermore National Laboratory[3] Salvat et. al., PENELOPE: An algorithm for Monte Carlo simulation of the penetration and energy loss of electrons and

positrons in matter, Nucl. Instr. and Meth. B 100 (1995) 31-46

1Department for Radiation Protection Physics, Institute for Nuclear and Particle Physics, TU Dresden

62

Calculation of Fluence Kerma Conversion Factors for Arbitrary Compounds

U. Reichelt, J. Henniger1

Estimating energy dose distributions of photon radia-tion fields is a common task in radiation protection andradiobiology. Often this can be done by estimating thekerma K. It is defined as the sum of kinetic energies ofall secondary particles per unit of mass, if the secondaryparticle equilibrium is fulfilled.Thereby, the spectral fluence ΦE or flux density ϕE isoften known from radiation transport simulations or ex-periments. If the effect of interest I (e.g. dose or kerma)is independent on the angle of incidence, it is possible touse fluence conversion factors fΦ→I to calculate I from

I(�r) =∫

ΦE(�r, E)fΦ→I(E) dE. (1)

This is an adequate assumption in photon dosimetry andthe dose can be derived by

fΦ→D = fΦ→K(1 − ηbr) (2)

with the fluence kerma conversion factor fΦ→K . Atlower photon energies the fraction of energy loss inducedby bremsstrahlung production of secondary particles ηbr

can be neglected. Thus the dose distribution is describedby the kerma distribution. The fluence kerma conversionfactor can be calculated from

fΦ→K(E) =1

ρ · Λ(E)W (3)

using the mean free path length Λ, mean energy loss perinteraction W and density of the material ρ.If the material consists of a single element, the meanenergy loss per interaction

W =∑

x

Px · W x (4)

is calculated by adding up the products of the condi-tional probability Px and the corresponding mean ki-netic energy loss W x for each interaction x. For photonradiation with photo effect (e), pair production (p), co-herent (no energy loss) and incoherent scattering (i) thisresults in

W = Pe ·E(1−ηx)+Pi · 〈Wi(µ)〉+Pp ·(E−2mec2). (5)

Therein µ denotes the cosine of the incoherent scatteringangle. For materials of low atomic number the X-ray lossηx is negligible. The expectation value of the inelasticscattering energy loss can be calculated from

〈Wi(µ)〉 =∫ +1

−1

f(µ) E(µ) dµ (6)

containing the probability density function (pdf) f(µ)and

E(µ) =E2

mec2(1 − µ)−1 + E. (7)

If the material consists of a mixture of ns elements withmolar amounts ν(s), and if binding effects are small, theconditional probabilities are expanded to

Px =ns∑

s=1

P (s)x (8)

with

P (s)x =

ν(s)∑ns

l=1 ν(l)· σ

(s)x∑

k σ(s)k

. (9)

Consequently (6) is changing to

〈Wi(µ)〉 =ns∑

s=1

[P

(s)i

Pi·∫ +1

−1

f(µ) · E2 · (1 − µ)mec2 + E(1 − µ)

dµ

]. (10)

Using equation (3), (8) and (9) it is possible to deter-mine the fluence kerma conversion factors also for nottabulated elemental compounds. The necessary data formean free path, interaction probability and pdf of theinelastic scattering angle can be calculated from severaldata libraries, here the EPDL [1] is used.Fig. 1 shows the achieved fluence kerma conversion fac-tors of water in comparison with values retrieved fromICRU 44 [2] and NISTIR 5632 [3]. The calculated graphfits very well with the data in the literature except a dis-crepancy at energy values above 20MeV. It is the con-sequence of bremsstrahlung emission by the secondaryparticles of the increasing pair production.Combined with fluence detectors in Monte-Carlo simula-tions the fluence kerma conversion factors are very usefulestimating dose distributions with high spatial resolu-tion.

Fig. 1 Comparison of fluence kerma conversion factors for

water; plots - self calculated, symbols - data from literature

[1] D.E. Cullen et al., Evaluated Photon Data Library, UCRL–50400, Vol. 6, Rev. 5 (1997)

[2] ICRU Report 44, Tissue Substitutes in Radiation Dosimetry and Measurement (1989)

[3] Technical Report NISTIR 5632, NIST, Gaithersburg, MD 20899 (1995)

1Department for Radiation Protection Physics, Institute for Nuclear and Particle Physics, TU Dresden

63

An Efficent Monte Carlo Detector for Fluence and Dose Estimationsfor Uncharged Particles

U. Reichelt, J. Henniger1

For treatment planning in brachytherapy the dose dis-tribution around a radioactive source has to be exactlyknown. Calculating this by means of Monte Carlo (MC)methods requires an adequate number of photon histo-ries.Some possible models are based on adding up all energydepositions in the detector elements. The low number ofinteractions in the lifetime of photons makes these ap-proaches less effective.Another possibility is to use a fluence detector. In do-ing so each crossed detector element at the photon pathcontributes to the detector response. If dose estimationsare requested, this fluence can be transformed by fluencedose conversion factors.The main problem in this detector type is to avoidthe calculation of interface passages, because this re-quires numerical instable or complex algorithms. This isachieved by means of the developed MC detector, whichis based on the fact that radiation transport is a Markovchain. This means that the interaction probabilities areindependent from all former events as long as the distri-bution parameters stay unchanged. At the way betweentwo interaction points this is fulfilled. Thus the energydependent mean free path length Λ(E) can be reducedto an arbitrary value Λmin(E) if after each allotted steplength a random number ξ ∈ (0; 1] is compared with theinteraction probability

Pi(E) =Λmin(E)

Λ(E). (1)

Only if ξ < Pi(E) is satisfied, a physical interaction isprocessed. Otherwise the photon is moving on withoutany changes (ongoing Markov chain).The end points of all resulting steps are used for thefluence estimation. Thereby, the particle weights nwgt

are added up to the detector element according to thediscretised spatial position (i, j, k)

Mijk ⇒ Mijk +nwgt

Λmin(E). (2)

Nevertheless, also the time and energy dependence canbe treated in a uniform manner. If the dose distributionis requested, the energy dependent conversion factorshave to be taken into account by

Mijk ⇒ Mijk +nwgt · fΦ→D(E)

Λmin(E). (3)

At the end of the photon transport simulation each de-tector element must be normalized to the number of sim-ulated photon histories and its specific volume.

Furthermore, the handling of material interface passagesin particle transport can be simplified if Λmin(E) is de-fined to be equal to or less than the smallest mean freepath of all occurring materials in the geometry [1]. So,there are no error-prone calculations of intercept pointswhen sampling the interaction length.As an example the calculation for an 125I brachyther-apy source by means of AMOS [2] is presented. For itsapplication in radiotherapy of prostate cancer the dosedistribution has to be well known with uncertainties lessthan 8%. Therefore, an annular detector of an axial sym-metric stack of 1000x 1000 hollow cylinders was used todetermine the dose in the surrounding water. The stackhas a radius of 5 cm and a length 10 cm. This results inan axial and radial binning of 0.005 cm using the mirrorsymmetry of the source. 125I emits photons of discreteenergy values in the range of 27.2 keV to 35.5 keV. Thusthe energy loss by bremsstrahlung can be neglected, andthe kerma is equivalent to the dose. Hence, the fluencekerma conversion factors obtained by [3] could be used.The simulation was done with a common personal com-puter.Figure 1 shows the dose distribution per emitted photon.The lower dose area on top of the brachy source origi-nates from self-shielding. Normally the distribution in adistance up to 5 cm is of therapeutic interest. In radialdirection results with statistical uncertainties below 5%are achievable up to 25 cm. Along this distance the doserate varies over nine orders of magnitude.This example proves the ability of the presented algo-rithm to simulate MC detectors of high spatial resolu-tion. The efficiency enables small statistical error mar-gins in simulations on state of the art personal comput-ers. This is required or at least advantageous for manyapplications in medicine and radiobiology.

Fig. 1 Dose distribution in water induced by an 125I

brachytherapy source

[1] I. Lux, Monte Carlo particle transport methods: neutron and photon calculation, CRC Press, Boca Raton (1991)

[2] J. Henniger, AMOS - ein multivalent nutzbares Programmsystem zur Berechnung von Strahlungstransportproblemen,Strahlenschutz: Physik und Messtechnik, Band 1, Verlag TUV Rheinland (1994) 145-150

[3] U. Reichelt, J. Henniger, Calculation of Fluence Kerma Conversion Factors, This Report p. 63

1Department for Radiation Protection Physics, Institute for Nuclear and Particle Physics, TU Dresden

64

In-Beam PET Imaging with LSO/APD-Array Detectors: First Results B, G, S

P. Crespo, M. Kapusta1, J. Pawelke, M.Moszynski

1W.Enghardt

The promising clinical results achieved at the heavy iontumor therapy project in GSI Darmstadt [1] promptedthe construction of a hospital-based facility at the Hei-delberg University Clinics. Here, delicate therapeuticsituations are expected to be treated with a rotatingbeam delivery (gantry) equipped with a dedicated in-beam PET system. The detectors of such a PET systemshould cover a large solid angle, be of small volume andminimal weight and insensitive to magnetic fields due tothe proximity to the beam bending magnets. The combi-nation of crystals of cerium-doped lutetium oxyorthosil-icate (LSO) and avalanche photodiode arrays (APDA)could solve this problem. While similar detectors haveshown their applicability to tracer imaging [2], their per-formance for in-beam PET is unknown up to now. Thus,they have to be experimentally studied before the finaldesign of the in-beam PET scanner for the Heidelbergtherapy facility. Therefore, we have investigated the per-formance and in-beam imaging capabilities of two posi-tion sensitive γ-ray detectors consisting of Hamamatsuavalanche photodiode arrays (S8550) individually cou-pled to crystals of cerium-doped lutetium oxyorthosili-cate (LSO). The two detectors were operated in coinci-dence at the GSI medical beam line.In a first set of experiments their imaging performancewas tested before, during and after the irradiationof phantoms of polymethylmethacrylate (PMMA) withcarbon ion beams with fluences equivalent to 1000 typ-ical daily therapeutic fractions. The detectors were po-sitioned at relatively small forward or backward angles(∼ 30◦) relative to the incoming radiation in order tosimulate the geometry of the detectors positioned at theedges of the tomograph planned for Heidelberg [3]. A68Ge line source placed vertically between the detectorswas imaged, with the source slightly shifted (14mm)towards one of the detectors in order not to be hitby the beam or by the flux of light particles leavingthe phantom that stopped the beam. Only minor en-ergy, time and spatial resolution deterioration was ob-served in the measurements performed under beam ir-radiation (Table 1 and Fig. 1), with the initial valuesbeing recovered after stopping the irradiation. Themean energy resolution obtained both with a line source(68Ge) and a point source (22Na) for the 64 channelsis 15.5± 0.4% FWHM, and the coincidence time reso-lution is 6.2± 0.2 ns FWHM. Both values are in agree-ment with our previous results [4] of 14.6± 0.5% FWHM(energy) and 3.0± 0.2 ns (time, extrapolated from mea-surement against BaF2) obtained for one single pixel.The slightly worse performance arises from the longerLSO crystals used (15mm vs. 10mm in [4]) and, more

importantly, from the increased number of channels in-troducing a higher time spread.A second set of experiments successfully imaged thedepth distribution of positron emitter radionuclides cre-ated in a phantom that stopped the high energy carbonion beam (Fig. 2). This set of imaging experiment isof importance since it simulates the application of suchdetectors for monitoring the 12C patient tumour irradi-ations. It checks furthermore the imaging capability ofthe detectors after being exposed to the back and for-ward cones of light particles resulting from high fluencesof 12C delivered to PMMA phantoms.The obtained results show that LSO is a suitable ma-terial for in-beam PET and that its coupling withavalanche photodiode arrays is feasible for a PET systemdedicated to in-beam monitoring of ion therapy.

Table 1 Performance of the LSO/APDA detectors.

ExperimentBeforeirradiation

Detectorsup-beam

Detectorsdown-beam

Afterirradiation

12C beam (AMeV) No 342.0 342.0 No

Phantom No Yes Yes No

∆E/E (fwhm, %) 15.5± 0.5 15.5± 0.3 16.4± 0.6 15.5± 0.3

∆t (fwhm, ns) 6.2± 0.2 6.1± 0.2 6.3± 0.2 6.2± 0.2

Fig. 1 Imaging a 68Ge line source under several irradiation

conditions.

Fig. 2 Positron emitter distribution generated by stopping

the carbon beam in a phantom. Imaged with the LSO/APDA

detectors. Intensity scale as in Fig. 1.

[1] D. Schulz-Ertner et al., Int. J. Radiation Oncology Biol. Phys., 53 (2002) 36

[2] B.J. Pichler et al., IEEE Trans. Nucl. Sci., 48 (2001) 1291

[3] P. Crespo et al., IEEE Trans. Nucl. Sci., submitted for publication

[4] M. Kapusta et al., Nucl. Inst. and Meth. A, 504 (2003) 139

1Andrzej Soltan Institute for Nuclear Studies, 05-400 Otwock-Swierk, Poland.

65

Experimental Investigations on In-Beam PET for Proton Therapy G, E

K. Parodi, F. Ponisch, W.Enghardt

Following the promising results of [1], further phantomexperiments have been performed at GSI in order toaddress the feasibility and accuracy of in-beam positronemission tomography (PET) for proton therapy. Forthe most challenging issue of range monitoring it hasbeen demonstrated for mono-energetic as well as forspread-out Bragg peak (SOBP) irradiation that rangedifferences of about 0.9 mm can be resolved (Fig. 1).The activation study of an inhomogenous phantom alsosupported the usefulness of PET for assessment of lo-cal tissue modifications or field misalignments leadingto strong density gradients in the beam path duringfractionated therapy. Despite the limitations of thepresently used calculation model for other materialsthan lucite [1] [2], a fairly good correspondence espe-cially with respect to the range spike after the lateraldensity gradient polyethylene/lung was observed in mea-sured and predicted images (Fig. 2). Finally, the valueof PET for lateral position verification was assessed byirradiation of a specially designed lucite phantom allow-ing insertion of radiographic films for the simultaneousmeasurement of the lateral dose distribution (Fig. 3).All these results, more deeply addressed in [2], stronglysupport the value of in-beam PET for proton therapy.However, the extension of such millimetre precision toreal clinical cases at least in low perfused tissues requiresa detailed description of all reaction channels leading to

0 50 100 150 200Penetration depth / mm

0.00.2

0.4

0.6

0.8

1.0

Act

ivity

/ a.

u.

171.62 MeV172.13 MeV172.63 MeV173.13 MeV

0 50 100 150 200Penetration depth / mm

0.00.2

0.4

0.6

0.8

1.0

Act

ivity

/ a.

u.

156.06 - 171.62 MeV156.60 - 172.13 MeV157.13 - 172.63 MeV157.66 - 173.13 MeV

Fig. 3 Measured depth profiles of β+-activity induced by

mono-energetic and SOBP (formed by 11 equally spaced en-

ergy steps) irradiation stopped in lucite at maximum depths

differing less than 1mm.

0 50 100 150Penetration depth / mm

-15-10-505

1015

Late

ral d

imen

sion

(m

m)

1 3 1 4 1 2 4 1

5

-15-10-505

1015

1 3 1 4 1 2 4 1

5

Fig. 2 Measured (top) and predicted (bottom) PET im-

ages induced by a 156.06 MeV pencil-like proton beam in a

target with inserts of polyethylene (1), lucite (2) and bone

(3), muscle (4) and lung (5) equivalent materials. The dif-

ferent regions are marked by the dashed lines.

β+-emitters in combination with an accurate knowledgeof the tissue stoichiometry for a realistic prediction to becompared with the PET measurement. Investigationson the influence of the 11C/15O ratio reported in [2] indi-cate also the sensitivity of the β+-activity distal edge tothe time course of the irradiation and PET acquisition.Hence, an optimal irradiation strategy in combinationwith proper off-line analysis should be considered inorder to achieve the required accuracy for range moni-toring. In perspective of the clinical implementation atthe ion beam facility at Heidelberg, the extension of themodeling to include more reaction channels in combina-tion with further PET experiments in pure phantom forvalidation of the available cross-section data is planned.

0 50 100 150 200 250Penetration depth / mm

5

10

15

20

25

Wid

th /

mm

PET meas.PET pred.Films meas.

Fig. 3 Lateral broadening (FWHM) of the measured (points)

and predicted (squares) β+-activity in comparison to the op-

tical density of the exposed films (stars) for a 190.37 MeV

proton irradiation. A vertical offset was introduced in order

to adjust the distributions to the same average FWHM of

the PET response.

[1] K. Parodi et al., Phys. Med. Biol. 47 (2002) 21

[2] K. Parodi, Ph.D. Thesis, in preparation (2004)

66

A New Computer System for In-Beam PET Data Evaluation G

F. Ponisch, W.Enghardt

The reconstruction of in-beam PET data acquired dur-ing the therapeutic irradiation with heavy ions at GSIDarmstadt requires high computational power since thePET images have to be available as shortly as possi-ble after finishing a fractionated irradiation, however,at the latest before applying the next daily dose frac-tion. Furthermore, the number of patients treated perday (up to 15) increases steadily and more sophisticatedand thus time consuming reconstruction techniques ase.g. the scatter correction introduced recently [1] andnew data evaluation tools [2] are applied. Therefore, wehave adapted our data processing software to a state-of-the-art computer system (Hewlett-Packard Linux clus-ter located at the computing centre of FZR). The clusterconsists of 40 CPU, from which up to 4 can be exclu-sively occupied for in-beam PET purposes and it hasbeen working for the routine PET data processing forheavy ion therapy since 2002. However, there remaintwo drawbacks: (i) The data transfer of the raw PET-data from the GSI to FZR and the retransfer of thereconstructed images to GSI requires a reliable long-distance data transmission. (ii) The effectiveness of thePET data evaluation correlates with the load of the localnetwork which is of course not as efficient like a directconnection to a PC.These disadvantages can be overcome by means of anon-site computer system placed at the therapy facilityof GSI Darmstadt. Such a system provides a better in-time data evaluation [2] since it simplifies the data han-dling and it fastens the interactive calculation of devia-tions between planned and applied dose. Thus, in 2003we have installed a new dual processor workstation (Fu-jitsu Siemens Celsius R610) with Intel Xeon processorsof 3.06GHz equipped with 4GB DDR-SDRAM mem-ory and a fast redundant SCSI RAID-1 storage disk ar-ray with a capacity of 122GB that is sufficient to storedata for future beam-times (each ≈ 5GB). The worksta-tion is connected via a Gigabit Ethernet LAN networkdevice to the intranet of the therapy facility and it isprotected against hacker attacks from outside using theGSI-firewall. The Linux operating system has been in-stalled since it provides a high level of transparency and

compatibility to the existing source codes and the shellscripts. Furthermore, the maintenance is rather easy andno additional license costs are expected for the future.Consequently, we have installed the SUSE Linux 8.2 withkernel 2.4.22, the IDL software package and the Intel C-Compiler. The benchmark tests below base on the com-parison of the C-code implementation of the maximumlikelihood tomographic reconstruction with scatter cor-rection [1] for PET data acquired during the carbon ionirradiation of a patient with a rather large PET image(118×100×116 voxels) and a large number of valid reg-istered coincidences (156800) according to the patientin [3]. Moreover, the influence of the different computersystems, compilers and the reconstruction parameters onthe run-time has been also studied. An example for areconstruction parameter of the scatter correction algo-rithm (single scatter simulation) affecting the run-timeis the lattice parameter (mean distance of the scatterpoints). Our first estimation assumes a lattice parame-ters of 10mm. Later investigations [4] showed that a sizeup to 15mm results in similar PET images. However, inorder to compare previous run-time tests with the recentcalculations, the smaller grid size is also presented. Theresults of the benchmark test are shown in Table 1 andmay summarized as follows: State-of-the-art high-endPC deliver the performance for a time efficient process-ing of sophisticated problems in scientific computing ona low cost level. The run-time is strongly dependent onthe applied C-compiler. The performance of the Intelcompiler on a Xeon CPU is about 1.5 times higher thanthat of the GNU compiler. The reconstruction code runsfaster on the Intel Xeon processor (3.06 GHz) than onthe tested workstation from IBM. Even the latest IBMmachine (1.45GHz Power4+CPU), whose clock cycle isby a factor of 4 higher than that of the processor shownin Table 1 is inferior to the new 3.06 GHz Xeon system.The measured run-time of 21min for the reconstructionof a typical PET image on the new Xeon workstation isless than the time needed for the treatment itself. There-fore, the two CPUs of a standard low cost workstationare sufficient to complete the data processing of the mea-sured and simulated PET data within less than 1/2hour.

Processor Clock cycle L2-Cache Reconstruction Time /min/GHz /kB Intel Compiler GNU Compiler

grid 10 mm grid 15 mm grid 15 mmIntel Xeon 2.0 256 97I 45I 71III

Intel Xeon 3.06 512 39II 21II 28IV

IBM Power3 0.375 8192 161V 81V -

Tab. 1 Comparison of the reconstruction time on different computers, reconstruction parameters and compilers; IVersion

6.0 with options: -O3 -tpp7 -xW, IIV7.1: -O3 -tpp7 -xW -ipo, IIIV2.95.2: -O3, IVV3.3: -march=pentium4, VV3.6.6: -O4.

[1] F. Ponisch, W. Enghardt, K. Lauckner, Phys. Med. Biol. 48 (2003) 2419[2] K. Parodi, W. Enghardt, Wiss.-Tech. Ber. FZR-372 (2002) 98[3] F. Ponisch, W. Enghardt, M. Schlett, Wiss.-Tech. Ber. FZR-372 (2002) 100[4] F. Ponisch, Wiss.-Tech. Ber. FZR-378 (2003) 61

67

In-Beam PET for Radiotherapy with Helium Beams: Model Predictions E, G

F. Fiedler, K. Parodi, W.Enghardt

For a first estimation on the amount and spatial dis-tribution of β+-emitters produced by helium beams intissue the approach of [1] was used. The experimen-tal cross sections from references [2-7] (Fig. 1) werecombined with the particle fluences calculated by theFLUKA-Simulation code [1], [2] to get the number ofβ+-emitters produced in a PMMA-target. For the reac-tions 16O(3He,3 Hen)15O and 16O(α, αn)15O the crosssections were estimated from the ratio corresponding re-actions with 12C and the 12C(p, p n)11C/16O(p, p n)15O,because there is a lack of experimental data. The pre-dicted β+-activity distributions resulting from the mostimportant positron emitters 11C and 15O produced in a9×9×20 cm3 PMMA-target by a 3He and 4He beam areshown in Figs. 3 and 4. The calculations were made for106 particles and the distributions are convolved withthe spatial resolution of the imaging system (FWHM =5 mm).

Fig. 1 Cross sections for the reactions 12C(α, αn)11 C resp.12C(3He,3 Hen)11C and 12C(3He, a)11C, the symbols mark

the data given in refs. [2-7].

Fig. 2 Stopping power of different ions in dependence on the

residual range in water.

For getting the same physical dose at the same parti-cle range with protons a 20 times higher particle fluencethan in the carbon case is required. Since the β+-emitterproduction rate per proton is about ten percent of the12C induced reactions and from this ratio an increase ofthe activity signal up to a factor of 2 was expected [10].Comparing the stopping power for carbon and helium(Fig. 2) we need a particle fluence about 6 times largerthan in the carbon case to get the same physical dose.The number of β+-emitters produced per incident par-ticle is about 10 times less than for carbon ions. Thusa signal approximately half as intense as for a carbonbeam is expected if the same physical dose is applied. If

comparing to the same biologically equivalent dose [11]a further increase of the helium induced signal up to amaximum of about 3 can be expected because of thehigher relative biological effectivness of carbon ions.

Ion E/AMeV Range/cm 15O 11C

1H 140 ∼ 12.0 12300 2665012C 270.55 ∼ 12.2 100400 2661003He 166 ∼ 12.2 10672 310224He 140 ∼ 12.1 6976 20361

Tab. 1 Prediction on the number of β+- emitters produced

in a 9 × 9 × 20 cm3 PMMA-target by helium beams with a

particle fluence of 10 6 for therapeutic energies in comparison

with the confirmed predictions for protons [10] and carbon

ions [10].

Fig. 3 Prediction of β+-emitter distribution for 11C and 15O

in case of a 3He-beam with E = 166 AMeV in a 9×9×20 cm3

PMMA-target.

Fig. 4 Prediction of β+-emitter distribution for 11C and15O in case of an alpha beam with E = 140 AMeV in a

9 × 9 × 20 cm3 PMMA-target.

[1] K. Parodi et al., Phys. Med. Biol. 47 (2002) 21[2] M. Lindner, R.N. Osborne, Phys. Rev. 91 (1953) 1501[3] W.E. Crandall et al., Phys. Rev. 101 (1956) 329[4] D.R.F.Cochran, J.D. Knight, Phys. Rev. 128 (1962)

1281[5] J. Radin, Phys. Rev. C 2 (1970) 793[6] R.L. Hahn, E. Ricci, Phys. Rev 146 (1966) 650[7] S.D.Cirilov et al., Nucl. Phys. 77 (1966) 472[8] A. Fasso et al., Proc. Monte Carlo 2000 Conference,

Lisbon, October 23-26 2000 (2001) 159[9] A. Fasso et al., Proc. Monte Carlo 2000 Conference,

Lisbon, October 23-26 2000 (2001) 955[10] K.Parodi et al., Phys. Med. Biol. 45 (2000) N151[11] M. Belli et al., Int. J. Rad. Biol. 76 (2000) 1095

68

The Prediction of the β+-Activity Distribution on the Basis of the TreatmentPlanning CT Using the FLUKA Simulation Code B, E, G

F.Fiedler, K. Parodi, A. Ferrari1, W. Enghardt

The in-beam PET technique for quality assurance of12C therapy is based on the comparison of β+-activitydistributions measured during therapeutic irradiationswith those predicted from the treatment plan. Thegeneralisation of this method to all species of thera-peutically relevant ions requires a precise descriptionof both the stopping process and the nuclear interac-tions of the projectiles in tissue. The FLUKA radiationtransport simulation code [1], [2] may meet these re-quirements since the implementation of suitable modelsfor describing light ion nuclear interactions is in progress.Furthermore, it has been demonstrated that measure-ments of proton beam induced β+-activity distributionscan be rather well reproduced by calculations based onthe FLUKA code [3]. The approach developed for theproton case may be feasible for other ion species: In afirst step the spatial distribution of the flux of primaryand secondary particles φ(E, A, Z, r) is calculated onthe basis of the FLUKA internal models of nuclear re-actions. In a second step a realistic description of theproduction of positron emitters is obtained by combin-ing these results with measured or semi-empirical crosssections [3].The β+-activity produced via target fragmentationstrongly depends on the stoichiometric tissue compo-sition, especially on the 16O/12C ratio, which varies be-tween 0.22 for adipose tissue and 4.05 for muscle. Thisratio determines the relative abundances of 15O (T1/2

= 2.03 min) and 11C (T1/2 = 20.38 min) as the domi-nating β+-active reaction products, and because of thedifferent half-lives the activity in different tissue types.Especially for projectiles with Z < 5, where β+-emittercan only be produced via target fragmentation, takinginto account the tissue stoichiometry is unavoidable fora correct description but also for heavier projectiles ase.g. 12C a refinement of the prediction of the β+-activity

Fig. 1 left: Patient CT, right: Profile of Hounsfield units

along the section AA′.

distribution from the treatment plan is expected. In atypical patient CT the Hounsfield numbers range from

−1024 to 3071 representing different densities and dif-ferent chemical compositions of tissue (Fig. 1). TheCT numbers are converted by a Fortran code into a filewhich is adopted by FLUKA as geometrical input. Forderiving the stoichiometric information we use the con-version of CT numbers to elemental weights establishedby [4],

Fig. 2 Chemical composition of the tissue along the section

AA′ in Fig. 1 referring to the segmentation described in the

text. Abundances below 1.5 % (Na, Mg, S, Ar, K) are not

displayed in this figure.

which gives the chemical composition for Hounsfieldnumbers between −1000 and 1600. Hounsfield unitslower than −950 are set to the composition of air, be-tween −120 and −950 the composition of lung is as-signed, in the range of soft tissue between −120 and 200a finer raster is applied and in the region above 200 agrid of 100 Hounsfield units is used. For Hounsfield unitslarger than 1600 the chemical composition has been ex-trapolated with the step width of 100 Hounsfield unitsup to 3000, the Hounsfield numbers larger than 3000are treated as titanium. This segmentation leads to 39different tissue compositions (Fig. 2). For each portalthe distribution of position emitters can be calculatedfor each pencil beam using a file containing general in-formation for the patient as well as the beam energy,focus and position as well as the number of particles forall points to be irradiated. This can be done before thetreatment starts, after the fractionated irradiation theseresults can be combined with the time protocol of theirradiation and give for each positron emitter the timewhen it was created. Then all processes from β+-decayto the detection of γ-rays can be done as described in [5].

[1] A. Fasso et al., Proc. Monte Carlo 2000Conference, Lisbon, October 23-26 2000 (2001) 159

[2] A. Fasso et al., Proc. Monte Carlo 2000Conference, Lisbon, October 23-26 2000 (2001) 955

[3] K. Parodi et al., Phys. Med. Biol. 47 (2002) 21[4] W. Schneider et al., Phys. Med. Biol. 45 (2000) 459[5] F. Ponisch, Wiss.-Tech. Ber. FZR-378 (2003)

1CERN, Geneva, Switzerland

69

Monte Carlo Simulations on In-Beam PET Imaging of Photon Radiotherapy:Target Activation via Photonuclear Reactions

H. Muller, W.Enghardt

Positron emission tomography (PET) plays an increas-ingly important role as a diagnostic method for support-ing radiation therapy [1]. Applying appropriate radio-tracers PET is capable of quantifying the metabolic ac-tivity of tissue and provides, therefore, information fora precise target definition during radiotherapeutic treat-ment planning and for monitoring the tumor responseduring and after the treatment. Furthermore, the com-bination of PET with radiotherapeutic treatment units(in-beam PET) offers a unique possibility of a three di-mensional in situ and non-invasive dose delivery moni-toring, which allows the irradiation field position to becontrolled and local dose deviations from the treatmentplan to be quantified [2]. This technique has been de-veloped for ion beam therapy [2, 3], but is expected tohave a considerable potential for improving the precisionof therapy with hard photon beams [1].

In this contribution the yield of positron emitters viathe (γ, n) reaction in various organic materials underthe irradiation with bremsstrahlung is simulated usingGEANT4 [4]. Calculations are carried out for electronbeams of 15, 18, 21 and 50 MeV, which hit a layer of1 mm tungsten and produce thus bremsstrahlung. Theelectrons are then absorbed in 5 cm iron. A parallel pho-ton beam is formed by randomly sampling the startingpoints of the photons within an area of 4 × 4 cm2. Thisbeam impinges centrally on targets of 9×9× 20 cm3 con-sisting of various materials. The resulting dose D andactivity A depth profiles as well as the ratio α = A/Dare shown in Fig. 1 for the case of 50 MeV electronsand target material PMMA. Here, the activity is de-rived from the number of produced 11C and 15O nucleiby taking into account their half lives of T1/2 = 20.38 mand T1/2 = 2.03 m, respectively. The dose distributionshows a broad maximum around 50 mm depth, while theactivity drops down continuously with increasing depth.It should be noted that like in the case of ion therapy theactivity distribution is not proportional to the dose dis-tribution. This concerns, however, mainly the entranceregion with its low dose delivery before reaching the sec-ondary electron equilibrium. From a certain depth onthe ration α of activity to dose remains nearly constant.

The ratio α averaged over the target region irradiatedby the 4 cm × 4 cm photon beam is gathered in Table 1for all energies and materials considered.

)-1 γD

( p

Gy

0.2

0.3

0.4

0.5

0.6

0.7

)-1 γ -3

A (

nB

q c

m2.22.42.62.8

33.23.43.63.8

)-3

cm

-1 (

kB

q G

yα

02468

101214161820

Depth (mm) 0 20 40 60 80 100 120 140 160 180 200

Fig. 1 Depth profile of dose D (upper panel), distribution

of activity A (middle panel) and the ratio α = A/D (lower

panel) in PMMA for 50 MeV photon beam.

Table 1 The ratio α ( BqGy−1 cm−3) of activity to dose for

various electron beam energies and target materials.

Energy(MeV) 15 18 21 50PMMA 21 113 359 5327Adipose 15 89 244 4177Bone 166 406 988 6163

Muscle 37 155 443 3631

The strong increase of α with energy for all materialsmakes the higher energies the better candidates for in-beam PET imaging. Comparing to the clinically applied12C and p beams where ratios of α ≈ 200 Bq Gy−1 cm−3

and α ≈ 600 Bq Gy−1 cm−3, respectively, were found weget for photons produced with electron energies above20 MeV an even higher activity per dose.

[1] A. Brahme, Acta Oncol. 42 (2003) 123

[2] W. Enghardt et al., Nucl. Instr. and Meth. A (in press)

[3] E. Urakabe et al., Jpn. J. Appl. Phys. 40 (2001) 2540

[4] S. Agostinelli et al., Nucl. Instr. and Meth. A506 (2003) 250

70

Monte Carlo Simulation on In-Beam PET Imaging of Photon Radiotherapy:Pair Production

H.Muller, W.Enghardt

In the previous contribution [1] we have discussedthe use of the positron emitters 11C and 15O pro-duced by bremsstrahlung during therapy for imag-ing on the basis of positron emission tomography(PET). In this contribution we discuss the possibilityof the direct use of the annihilation radiation follow-ing e+e−-pair production during photon therapy forPET imaging simultaneously to therapeutic irradiations.

Depth ( mm )0 20 40 60 80 100 120 140 160 180 200

)-3

cm

-1 (

Gy

δ

0

10

20

30

40

50

60

70

80

710×

Fig. 1 The ratio δ of the number of pair production positrons

to dose in PMMA for 50 MeV photon beam.

In Fig. 1 the distribution of the ratio of the number ofproduced positrons to the dose is depicted. As in the pre-vious contribution the proportionality to dose delivery isstrongly violated in the entrance region of the target.

A second problem concerns the question whether it ispossible to image anatomic structures by measuringpositrons from pair production in a positron camerato get a method for controlling the positioning of thepatients. Since these positrons have high initial ener-gies they may travel appreciable distances before theyannihilate at rest. This effect degrades the resolutionof the image. The distances between the space pointsof positron creation and annihilation for the case of abremsstrahlung spectrum created by 50 MeV electronsunder conditions described in [1] are shown in Fig. 2.Since the positrons are mainly produced in forward di-rection, we distinguish between the projections paralleland transverse to the direction of the incoming beam.

)-1 γ -1E

ven

ts (

mm

0

0.005

0.01

-10 0 10 20

Muscle

-10 0 10 20

Adipose

Distance ( mm )-10 0 10 20

RMS/mm

X: 4.62

Z: 10.91

RMS/mm

X: 6.90

Z: 12.74

RMS/mm

X: 7.29

Z: 13.05

Bone

Fig. 2 Distance between production and annihilation trans-

verse -X- (full histograms) and longitudinal -Z- (dashed his-

tograms) to the beam direction for various types of tissue

indicated in the figure for 50 MeV photon beam.

As expected the transverse distribution is symmetric to

zero and much smaller than the longitudinal distribu-tion, which shows a strong asymmetry.In Fig. 3 the distribution of annihilation points ofpositrons in a target filled with water is shown. In itscentre a cube of 1 cm side length is located containingdifferent kinds of tissue. From the different density ofannihilation points in the various materials can be con-cluded that, in spite of the broad distributions of Fig. 2,an imaging on the basis of PET should be possible. Es-pecially perpendicular to the beam direction the positionof steep density gradients can be localized.

H2O

_Adi

pose )

-1 γ

Eve

nts

(

0

0.05

0.1

0.15-410×

H2O

_Air

0

0.05

0.1

-410×

H2O

_Bon

e

0

0.1

0.2

-410×

H2O

_Mus

cle

0

0.05

0.1

0.15

-410×

x ( mm )-20 -10 0 10 20

z ( mm )90 100 110 120

Fig. 3 Distribution of annihilation points in a water phan-

tom of 9× 9× 20 cm3 containing in its center a cube of 1 cm

side length filled with different types of tissue as indicated

in the figure for 50 MeV photon beam.

The production rate of positrons is estimated accordingto the following assumptions:

- The number density of positrons per dose producedin PMMA amounts to about δ ≈ 2×108 Gy−1cm−3

(see Fig. 1).- For the volume the whole target is taken

V = 4 × 4 × 20 cm3 = 320 cm3.- A dose of D = 1 Gy is delivered in t = 100 s.- For the duty cycle a value of f = 10−3 is taken

typical for medical linear accelerators.- This results in an annihilation γ-ray flux of Nγ =

2 δ V D/(f t) ≈ 12×1011s−1 and a total γ-ray num-ber of 12 × 1010.

- For a single detector of area A = 4× 4 mm2 placedin a distance of R = 400 mm we get a counting rateof NA = Nγ A / (4π R2) ≈ 107 s−1 during themacro pulse of the electron beam.

PET imaging of annihilation γ-rays following pair pro-duction is thus not feasible because of overloading thePET detectors during the beam macro pulses of medicalaccelerators, whereas target activation should be mea-surable during the pauses between the macro pulses.

[1] H. Muller, W. Enghardt, This Report, p. 70

71